char *Loc::toChars() { OutBuffer buf; if (filename) { buf.printf("%s", filename); } if (linnum) buf.printf("(%d)", linnum); buf.writeByte(0); return (char *)buf.extractData(); }
void LibOMF::write() { if (global.params.verbose) fprintf(global.stdmsg, "library %s\n", libfile->name->toChars()); OutBuffer libbuf; WriteLibToBuffer(&libbuf); // Transfer image to file libfile->setbuffer(libbuf.data, libbuf.offset); libbuf.extractData(); FileName::ensurePathToNameExists(libfile->name->toChars()); libfile->writev(); }
char *ModuleDeclaration::toChars() { OutBuffer buf; if (packages && packages->dim) { for (size_t i = 0; i < packages->dim; i++) { Identifier *pid = packages->tdata()[i]; buf.writestring(pid->toChars()); buf.writeByte('.'); } } buf.writestring(id->toChars()); buf.writeByte(0); return (char *)buf.extractData(); }
void Library::write() { if (global.params.verbose) printf("library %s\n", libfile->name->toChars()); OutBuffer libbuf; WriteLibToBuffer(&libbuf); // Transfer image to file libfile->setbuffer(libbuf.data, libbuf.offset); libbuf.extractData(); char *p = FileName::path(libfile->name->toChars()); FileName::ensurePathExists(p); //mem.free(p); libfile->writev(); }
void Module::parse() #endif { char *srcname; unsigned char *buf; unsigned buflen; unsigned le; unsigned bom; //printf("Module::parse()\n"); srcname = srcfile->name->toChars(); //printf("Module::parse(srcname = '%s')\n", srcname); buf = srcfile->buffer; buflen = srcfile->len; if (buflen >= 2) { /* Convert all non-UTF-8 formats to UTF-8. * BOM : http://www.unicode.org/faq/utf_bom.html * 00 00 FE FF UTF-32BE, big-endian * FF FE 00 00 UTF-32LE, little-endian * FE FF UTF-16BE, big-endian * FF FE UTF-16LE, little-endian * EF BB BF UTF-8 */ bom = 1; // assume there's a BOM if (buf[0] == 0xFF && buf[1] == 0xFE) { if (buflen >= 4 && buf[2] == 0 && buf[3] == 0) { // UTF-32LE le = 1; Lutf32: OutBuffer dbuf; unsigned *pu = (unsigned *)(buf); unsigned *pumax = &pu[buflen / 4]; if (buflen & 3) { error("odd length of UTF-32 char source %u", buflen); fatal(); } dbuf.reserve(buflen / 4); for (pu += bom; pu < pumax; pu++) { unsigned u; u = le ? readlongLE(pu) : readlongBE(pu); if (u & ~0x7F) { if (u > 0x10FFFF) { error("UTF-32 value %08x greater than 0x10FFFF", u); fatal(); } dbuf.writeUTF8(u); } else dbuf.writeByte(u); } dbuf.writeByte(0); // add 0 as sentinel for scanner buflen = dbuf.offset - 1; // don't include sentinel in count buf = (unsigned char *) dbuf.extractData(); } else { // UTF-16LE (X86) // Convert it to UTF-8 le = 1; Lutf16: OutBuffer dbuf; unsigned short *pu = (unsigned short *)(buf); unsigned short *pumax = &pu[buflen / 2]; if (buflen & 1) { error("odd length of UTF-16 char source %u", buflen); fatal(); } dbuf.reserve(buflen / 2); for (pu += bom; pu < pumax; pu++) { unsigned u; u = le ? readwordLE(pu) : readwordBE(pu); if (u & ~0x7F) { if (u >= 0xD800 && u <= 0xDBFF) { unsigned u2; if (++pu > pumax) { error("surrogate UTF-16 high value %04x at EOF", u); fatal(); } u2 = le ? readwordLE(pu) : readwordBE(pu); if (u2 < 0xDC00 || u2 > 0xDFFF) { error("surrogate UTF-16 low value %04x out of range", u2); fatal(); } u = (u - 0xD7C0) << 10; u |= (u2 - 0xDC00); } else if (u >= 0xDC00 && u <= 0xDFFF) { error("unpaired surrogate UTF-16 value %04x", u); fatal(); } else if (u == 0xFFFE || u == 0xFFFF) { error("illegal UTF-16 value %04x", u); fatal(); } dbuf.writeUTF8(u); } else dbuf.writeByte(u); } dbuf.writeByte(0); // add 0 as sentinel for scanner buflen = dbuf.offset - 1; // don't include sentinel in count buf = (unsigned char *) dbuf.extractData(); } } else if (buf[0] == 0xFE && buf[1] == 0xFF) { // UTF-16BE le = 0; goto Lutf16; } else if (buflen >= 4 && buf[0] == 0 && buf[1] == 0 && buf[2] == 0xFE && buf[3] == 0xFF) { // UTF-32BE le = 0; goto Lutf32; } else if (buflen >= 3 && buf[0] == 0xEF && buf[1] == 0xBB && buf[2] == 0xBF) { // UTF-8 buf += 3; buflen -= 3; } else { /* There is no BOM. Make use of Arcane Jill's insight that * the first char of D source must be ASCII to * figure out the encoding. */ bom = 0; if (buflen >= 4) { if (buf[1] == 0 && buf[2] == 0 && buf[3] == 0) { // UTF-32LE le = 1; goto Lutf32; } else if (buf[0] == 0 && buf[1] == 0 && buf[2] == 0) { // UTF-32BE le = 0; goto Lutf32; } } if (buflen >= 2) { if (buf[1] == 0) { // UTF-16LE le = 1; goto Lutf16; } else if (buf[0] == 0) { // UTF-16BE le = 0; goto Lutf16; } } // It's UTF-8 if (buf[0] >= 0x80) { error("source file must start with BOM or ASCII character, not \\x%02X", buf[0]); fatal(); } } } #ifdef IN_GCC // dump utf-8 encoded source if (dump_source) { // %% srcname could contain a path ... d_gcc_dump_source(srcname, "utf-8", buf, buflen); } #endif /* If it starts with the string "Ddoc", then it's a documentation * source file. */ if (buflen >= 4 && memcmp(buf, "Ddoc", 4) == 0) { comment = buf + 4; isDocFile = 1; return; } if (isHtml) { OutBuffer *dbuf = new OutBuffer(); Html h(srcname, buf, buflen); h.extractCode(dbuf); buf = dbuf->data; buflen = dbuf->offset; #ifdef IN_GCC // dump extracted source if (dump_source) d_gcc_dump_source(srcname, "d.utf-8", buf, buflen); #endif } #if IN_LLVM Parser p(this, buf, buflen, gen_docs); #else Parser p(this, buf, buflen, docfile != NULL); #endif p.nextToken(); members = p.parseModule(); md = p.md; numlines = p.loc.linnum; DsymbolTable *dst; if (md) { this->ident = md->id; dst = Package::resolve(md->packages, &this->parent, NULL); } else { dst = modules; /* Check to see if module name is a valid identifier */ if (!Lexer::isValidIdentifier(this->ident->toChars())) error("has non-identifier characters in filename, use module declaration instead"); } // Update global list of modules if (!dst->insert(this)) { Dsymbol *prev = dst->lookup(ident); assert(prev); Module *mprev = prev->isModule(); if (mprev) error(loc, "from file %s conflicts with another module %s from file %s", srcname, mprev->toChars(), mprev->srcfile->toChars()); else { Package *pkg = prev->isPackage(); assert(pkg); error(loc, "from file %s conflicts with package name %s", srcname, pkg->toChars()); } } else { amodules.push(this); } }
Module *Module::load(Loc loc, Array *packages, Identifier *ident) { Module *m; char *filename; //printf("Module::load(ident = '%s')\n", ident->toChars()); // Build module filename by turning: // foo.bar.baz // into: // foo\bar\baz filename = ident->toChars(); if (packages && packages->dim) { OutBuffer buf; int i; for (i = 0; i < packages->dim; i++) { Identifier *pid = (Identifier *)packages->data[i]; buf.writestring(pid->toChars()); #if _WIN32 buf.writeByte('\\'); #else buf.writeByte('/'); #endif } buf.writestring(filename); buf.writeByte(0); filename = (char *)buf.extractData(); } m = new Module(filename, ident, 0, 0); m->loc = loc; /* Search along global.path for .di file, then .d file. */ char *result = NULL; FileName *fdi = FileName::forceExt(filename, global.hdr_ext); FileName *fd = FileName::forceExt(filename, global.mars_ext); char *sdi = fdi->toChars(); char *sd = fd->toChars(); if (FileName::exists(sdi)) result = sdi; else if (FileName::exists(sd)) result = sd; else if (FileName::absolute(filename)) ; else if (!global.path) ; else { for (size_t i = 0; i < global.path->dim; i++) { char *p = (char *)global.path->data[i]; char *n = FileName::combine(p, sdi); if (FileName::exists(n)) { result = n; break; } mem.free(n); n = FileName::combine(p, sd); if (FileName::exists(n)) { result = n; break; } mem.free(n); } } if (result) m->srcfile = new File(result); if (global.params.verbose) { printf("import "); if (packages) { for (size_t i = 0; i < packages->dim; i++) { Identifier *pid = (Identifier *)packages->data[i]; printf("%s.", pid->toChars()); } } printf("%s\t(%s)\n", ident->toChars(), m->srcfile->toChars()); } m->read(loc); m->parse(); #ifdef IN_GCC d_gcc_magic_module(m); #endif return m; }
void obj_write_deferred(Library *library) { for (int i = 0; i < obj_symbols_towrite.dim; i++) { Dsymbol *s = (Dsymbol *)obj_symbols_towrite.data[i]; Module *m = s->getModule(); char *mname; if (m) { mname = m->srcfile->toChars(); lastmname = mname; } else { //mname = s->ident->toChars(); mname = lastmname; assert(mname); } obj_start(mname); static int count; count++; // sequence for generating names /* Create a module that's a doppelganger of m, with just * enough to be able to create the moduleinfo. */ OutBuffer idbuf; idbuf.printf("%s.%d", m ? m->ident->toChars() : mname, count); char *idstr = idbuf.toChars(); idbuf.data = NULL; Identifier *id = new Identifier(idstr, TOKidentifier); Module *md = new Module(mname, id, 0, 0); md->members = new Array(); md->members->push(s); // its only 'member' is s if (m) { md->doppelganger = 1; // identify this module as doppelganger md->md = m->md; md->aimports.push(m); // it only 'imports' m md->massert = m->massert; md->marray = m->marray; } md->genobjfile(0); /* Set object file name to be source name with sequence number, * as mangled symbol names get way too long. */ char *fname = FileName::removeExt(mname); OutBuffer namebuf; unsigned hash = 0; for (char *p = s->toChars(); *p; p++) hash += *p; namebuf.printf("%s_%x_%x.%s", fname, count, hash, global.obj_ext); namebuf.writeByte(0); mem.free(fname); fname = (char *)namebuf.extractData(); //printf("writing '%s'\n", fname); File *objfile = new File(fname); obj_end(library, objfile); } obj_symbols_towrite.dim = 0; }
// Split a path into an Array of paths Strings *FileName::splitPath(const char *path) { char c = 0; // unnecessary initializer is for VC /W4 const char *p; OutBuffer buf; Strings *array; array = new Strings(); if (path) { p = path; do { char instring = 0; while (isspace((unsigned char)*p)) // skip leading whitespace p++; buf.reserve(strlen(p) + 1); // guess size of path for (; ; p++) { c = *p; switch (c) { case '"': instring ^= 1; // toggle inside/outside of string continue; #if MACINTOSH case ',': #endif #if _WIN32 case ';': #endif #if POSIX case ':': #endif p++; break; // note that ; cannot appear as part // of a path, quotes won't protect it case 0x1A: // ^Z means end of file case 0: break; case '\r': continue; // ignore carriage returns #if POSIX case '~': buf.writestring(getenv("HOME")); continue; #endif #if 0 case ' ': case '\t': // tabs in filenames? if (!instring) // if not in string break; // treat as end of path #endif default: buf.writeByte(c); continue; } break; } if (buf.offset) // if path is not empty { buf.writeByte(0); // to asciiz array->push(buf.extractData()); } } while (c); } return array; }
int runLINK() { #if _WIN32 if (global.params.is64bit) { OutBuffer cmdbuf; cmdbuf.writestring("/NOLOGO "); for (size_t i = 0; i < global.params.objfiles->dim; i++) { if (i) cmdbuf.writeByte(' '); const char *p = (*global.params.objfiles)[i]; const char *basename = FileName::removeExt(FileName::name(p)); const char *ext = FileName::ext(p); if (ext && !strchr(basename, '.')) // Write name sans extension (but not if a double extension) writeFilename(&cmdbuf, p, ext - p - 1); else writeFilename(&cmdbuf, p); FileName::free(basename); } if (global.params.resfile) { cmdbuf.writeByte(' '); writeFilename(&cmdbuf, global.params.resfile); } cmdbuf.writeByte(' '); if (global.params.exefile) { cmdbuf.writestring("/OUT:"); writeFilename(&cmdbuf, global.params.exefile); } else { /* Generate exe file name from first obj name. * No need to add it to cmdbuf because the linker will default to it. */ const char *n = (*global.params.objfiles)[0]; n = FileName::name(n); global.params.exefile = (char *)FileName::forceExt(n, "exe"); } // Make sure path to exe file exists FileName::ensurePathToNameExists(global.params.exefile); cmdbuf.writeByte(' '); if (global.params.mapfile) { cmdbuf.writestring("/MAP:"); writeFilename(&cmdbuf, global.params.mapfile); } else if (global.params.map) { const char *fn = FileName::forceExt(global.params.exefile, "map"); const char *path = FileName::path(global.params.exefile); const char *p; if (path[0] == '\0') p = FileName::combine(global.params.objdir, fn); else p = fn; cmdbuf.writestring("/MAP:"); writeFilename(&cmdbuf, p); } for (size_t i = 0; i < global.params.libfiles->dim; i++) { cmdbuf.writeByte(' '); cmdbuf.writestring("/DEFAULTLIB:"); writeFilename(&cmdbuf, (*global.params.libfiles)[i]); } if (global.params.deffile) { cmdbuf.writeByte(' '); cmdbuf.writestring("/DEF:"); writeFilename(&cmdbuf, global.params.deffile); } if (global.params.symdebug) { cmdbuf.writeByte(' '); cmdbuf.writestring("/DEBUG"); } if (global.params.dll) { cmdbuf.writeByte(' '); cmdbuf.writestring("/DLL"); } for (size_t i = 0; i < global.params.linkswitches->dim; i++) { cmdbuf.writeByte(' '); cmdbuf.writestring((*global.params.linkswitches)[i]); } /* Append the path to the VC lib files, and then the SDK lib files */ const char *vcinstalldir = getenv("VCINSTALLDIR"); if (vcinstalldir) { cmdbuf.writestring(" \"/LIBPATH:"); cmdbuf.writestring(vcinstalldir); cmdbuf.writestring("lib\\amd64\""); } const char *windowssdkdir = getenv("WindowsSdkDir"); if (windowssdkdir) { cmdbuf.writestring(" \"/LIBPATH:"); cmdbuf.writestring(windowssdkdir); cmdbuf.writestring("lib\\x64\""); } char *p = cmdbuf.toChars(); const char *lnkfilename = NULL; size_t plen = strlen(p); if (plen > 7000) { lnkfilename = FileName::forceExt(global.params.exefile, "lnk"); File flnk(lnkfilename); flnk.setbuffer(p, plen); flnk.ref = 1; if (flnk.write()) error(Loc(), "error writing file %s", lnkfilename); if (strlen(lnkfilename) < plen) sprintf(p, "@%s", lnkfilename); } char *linkcmd = getenv("LINKCMD64"); if (!linkcmd) { if (vcinstalldir) { OutBuffer linkcmdbuf; linkcmdbuf.writestring(vcinstalldir); linkcmdbuf.writestring("bin\\amd64\\link"); linkcmd = linkcmdbuf.toChars(); linkcmdbuf.extractData(); } else linkcmd = "link"; } int status = executecmd(linkcmd, p, 1); if (lnkfilename) { remove(lnkfilename); FileName::free(lnkfilename); } return status; } else { OutBuffer cmdbuf; global.params.libfiles->push("user32"); global.params.libfiles->push("kernel32"); for (size_t i = 0; i < global.params.objfiles->dim; i++) { if (i) cmdbuf.writeByte('+'); const char *p = (*global.params.objfiles)[i]; const char *basename = FileName::removeExt(FileName::name(p)); const char *ext = FileName::ext(p); if (ext && !strchr(basename, '.')) // Write name sans extension (but not if a double extension) writeFilename(&cmdbuf, p, ext - p - 1); else writeFilename(&cmdbuf, p); FileName::free(basename); } cmdbuf.writeByte(','); if (global.params.exefile) writeFilename(&cmdbuf, global.params.exefile); else { /* Generate exe file name from first obj name. * No need to add it to cmdbuf because the linker will default to it. */ const char *n = (*global.params.objfiles)[0]; n = FileName::name(n); global.params.exefile = (char *)FileName::forceExt(n, "exe"); } // Make sure path to exe file exists FileName::ensurePathToNameExists(global.params.exefile); cmdbuf.writeByte(','); if (global.params.mapfile) writeFilename(&cmdbuf, global.params.mapfile); else if (global.params.map) { const char *fn = FileName::forceExt(global.params.exefile, "map"); const char *path = FileName::path(global.params.exefile); const char *p; if (path[0] == '\0') p = FileName::combine(global.params.objdir, fn); else p = fn; writeFilename(&cmdbuf, p); } else cmdbuf.writestring("nul"); cmdbuf.writeByte(','); for (size_t i = 0; i < global.params.libfiles->dim; i++) { if (i) cmdbuf.writeByte('+'); writeFilename(&cmdbuf, (*global.params.libfiles)[i]); } if (global.params.deffile) { cmdbuf.writeByte(','); writeFilename(&cmdbuf, global.params.deffile); } /* Eliminate unnecessary trailing commas */ while (1) { size_t i = cmdbuf.offset; if (!i || cmdbuf.data[i - 1] != ',') break; cmdbuf.offset--; } if (global.params.resfile) { cmdbuf.writestring("/RC:"); writeFilename(&cmdbuf, global.params.resfile); } if (global.params.map || global.params.mapfile) cmdbuf.writestring("/m"); #if 0 if (debuginfo) cmdbuf.writestring("/li"); if (codeview) { cmdbuf.writestring("/co"); if (codeview3) cmdbuf.writestring(":3"); } #else if (global.params.symdebug) cmdbuf.writestring("/co"); #endif cmdbuf.writestring("/noi"); for (size_t i = 0; i < global.params.linkswitches->dim; i++) { cmdbuf.writestring((*global.params.linkswitches)[i]); } cmdbuf.writeByte(';'); char *p = cmdbuf.toChars(); const char *lnkfilename = NULL; size_t plen = strlen(p); if (plen > 7000) { lnkfilename = FileName::forceExt(global.params.exefile, "lnk"); File flnk(lnkfilename); flnk.setbuffer(p, plen); flnk.ref = 1; if (flnk.write()) error(Loc(), "error writing file %s", lnkfilename); if (strlen(lnkfilename) < plen) sprintf(p, "@%s", lnkfilename); } char *linkcmd = getenv("LINKCMD"); if (!linkcmd) linkcmd = "link"; int status = executecmd(linkcmd, p, 1); if (lnkfilename) { remove(lnkfilename); FileName::free(lnkfilename); } return status; } #elif linux || __APPLE__ || __FreeBSD__ || __OpenBSD__ || __sun pid_t childpid; int status; // Build argv[] Strings argv; const char *cc = getenv("CC"); if (!cc) cc = "gcc"; argv.push((char *)cc); argv.insert(1, global.params.objfiles); #if __APPLE__ // If we are on Mac OS X and linking a dynamic library, // add the "-dynamiclib" flag if (global.params.dll) argv.push((char *) "-dynamiclib"); #elif linux || __FreeBSD__ || __OpenBSD__ || __sun if (global.params.dll) argv.push((char *) "-shared"); #endif // None of that a.out stuff. Use explicit exe file name, or // generate one from name of first source file. argv.push((char *)"-o"); if (global.params.exefile) { argv.push(global.params.exefile); } else { // Generate exe file name from first obj name const char *n = (*global.params.objfiles)[0]; char *ex; n = FileName::name(n); const char *e = FileName::ext(n); if (e) { e--; // back up over '.' ex = (char *)mem.malloc(e - n + 1); memcpy(ex, n, e - n); ex[e - n] = 0; // If generating dll then force dll extension if (global.params.dll) ex = (char *)FileName::forceExt(ex, global.dll_ext); } else ex = (char *)"a.out"; // no extension, so give up argv.push(ex); global.params.exefile = ex; } // Make sure path to exe file exists FileName::ensurePathToNameExists(global.params.exefile); if (global.params.symdebug) argv.push((char *)"-g"); if (global.params.is64bit) argv.push((char *)"-m64"); else argv.push((char *)"-m32"); if (global.params.map || global.params.mapfile) { argv.push((char *)"-Xlinker"); #if __APPLE__ argv.push((char *)"-map"); #else argv.push((char *)"-Map"); #endif if (!global.params.mapfile) { const char *fn = FileName::forceExt(global.params.exefile, "map"); const char *path = FileName::path(global.params.exefile); const char *p; if (path[0] == '\0') p = FileName::combine(global.params.objdir, fn); else p = fn; global.params.mapfile = (char *)p; } argv.push((char *)"-Xlinker"); argv.push(global.params.mapfile); } if (0 && global.params.exefile) { /* This switch enables what is known as 'smart linking' * in the Windows world, where unreferenced sections * are removed from the executable. It eliminates unreferenced * functions, essentially making a 'library' out of a module. * Although it is documented to work with ld version 2.13, * in practice it does not, but just seems to be ignored. * Thomas Kuehne has verified that it works with ld 2.16.1. * BUG: disabled because it causes exception handling to fail * because EH sections are "unreferenced" and elided */ argv.push((char *)"-Xlinker"); argv.push((char *)"--gc-sections"); } for (size_t i = 0; i < global.params.linkswitches->dim; i++) { char *p = (*global.params.linkswitches)[i]; if (!p || !p[0] || !(p[0] == '-' && (p[1] == 'l' || p[1] == 'L'))) // Don't need -Xlinker if switch starts with -l or -L. // Eliding -Xlinker is significant for -L since it allows our paths // to take precedence over gcc defaults. argv.push((char *)"-Xlinker"); argv.push(p); } /* Add each library, prefixing it with "-l". * The order of libraries passed is: * 1. any libraries passed with -L command line switch * 2. libraries specified on the command line * 3. libraries specified by pragma(lib), which were appended * to global.params.libfiles. * 4. standard libraries. */ for (size_t i = 0; i < global.params.libfiles->dim; i++) { char *p = (*global.params.libfiles)[i]; size_t plen = strlen(p); if (plen > 2 && p[plen - 2] == '.' && p[plen -1] == 'a') argv.push(p); else { char *s = (char *)mem.malloc(plen + 3); s[0] = '-'; s[1] = 'l'; memcpy(s + 2, p, plen + 1); argv.push(s); } } /* Standard libraries must go after user specified libraries * passed with -l. */ const char *libname = (global.params.symdebug) ? global.params.debuglibname : global.params.defaultlibname; size_t slen = strlen(libname); if (slen) { char *buf = (char *)malloc(3 + slen + 1); strcpy(buf, "-l"); /* Use "-l:libname.a" if the library name is complete */ if (slen > 3 + 2 && memcmp(libname, "lib", 3) == 0 && (memcmp(libname + slen - 2, ".a", 2) == 0 || memcmp(libname + slen - 3, ".so", 3) == 0) ) { strcat(buf, ":"); } strcat(buf, libname); argv.push(buf); // turns into /usr/lib/libphobos2.a } #ifdef __sun argv.push((char *)"-mt"); #endif // argv.push((void *)"-ldruntime"); argv.push((char *)"-lpthread"); argv.push((char *)"-lm"); #if linux && DMDV2 // Changes in ld for Ubuntu 11.10 require this to appear after phobos2 argv.push((char *)"-lrt"); #endif if (!global.params.quiet || global.params.verbose) { // Print it for (size_t i = 0; i < argv.dim; i++) printf("%s ", argv[i]); printf("\n"); fflush(stdout); } argv.push(NULL); // set up pipes int fds[2]; if (pipe(fds) == -1) { perror("Unable to create pipe to linker"); return -1; } childpid = fork(); if (childpid == 0) { // pipe linker stderr to fds[0] dup2(fds[1], STDERR_FILENO); close(fds[0]); execvp(argv[0], argv.tdata()); perror(argv[0]); // failed to execute return -1; } else if (childpid == -1) { perror("Unable to fork"); return -1; } close(fds[1]); const int nme = findNoMainError(fds[0]); waitpid(childpid, &status, 0); if (WIFEXITED(status)) { status = WEXITSTATUS(status); if (status) { if (nme == -1) { perror("Error with the linker pipe"); return -1; } else { printf("--- errorlevel %d\n", status); if (nme == 1) error(Loc(), "no main function specified"); } } } else if (WIFSIGNALED(status)) { printf("--- killed by signal %d\n", WTERMSIG(status)); status = 1; } return status; #else printf ("Linking is not yet supported for this version of DMD.\n"); return -1; #endif }
Expression *BinExp::arrayOp(Scope *sc) { //printf("BinExp::arrayOp() %s\n", toChars()); if (type->toBasetype()->nextOf()->toBasetype()->ty == Tvoid) { error("Cannot perform array operations on void[] arrays"); return new ErrorExp(); } Expressions *arguments = new Expressions(); /* The expression to generate an array operation for is mangled * into a name to use as the array operation function name. * Mangle in the operands and operators in RPN order, and type. */ OutBuffer buf; buf.writestring("_array"); buildArrayIdent(&buf, arguments); buf.writeByte('_'); /* Append deco of array element type */ #if DMDV2 buf.writestring(type->toBasetype()->nextOf()->toBasetype()->mutableOf()->deco); #else buf.writestring(type->toBasetype()->nextOf()->toBasetype()->deco); #endif size_t namelen = buf.offset; buf.writeByte(0); char *name = (char *)buf.extractData(); /* Look up name in hash table */ StringValue *sv = arrayfuncs.update(name, namelen); FuncDeclaration *fd = (FuncDeclaration *)sv->ptrvalue; if (!fd) { /* Some of the array op functions are written as library functions, * presumably to optimize them with special CPU vector instructions. * List those library functions here, in alpha order. */ static const char *libArrayopFuncs[] = { "_arrayExpSliceAddass_a", "_arrayExpSliceAddass_d", // T[]+=T "_arrayExpSliceAddass_f", // T[]+=T "_arrayExpSliceAddass_g", "_arrayExpSliceAddass_h", "_arrayExpSliceAddass_i", "_arrayExpSliceAddass_k", "_arrayExpSliceAddass_s", "_arrayExpSliceAddass_t", "_arrayExpSliceAddass_u", "_arrayExpSliceAddass_w", "_arrayExpSliceDivass_d", // T[]/=T "_arrayExpSliceDivass_f", // T[]/=T "_arrayExpSliceMinSliceAssign_a", "_arrayExpSliceMinSliceAssign_d", // T[]=T-T[] "_arrayExpSliceMinSliceAssign_f", // T[]=T-T[] "_arrayExpSliceMinSliceAssign_g", "_arrayExpSliceMinSliceAssign_h", "_arrayExpSliceMinSliceAssign_i", "_arrayExpSliceMinSliceAssign_k", "_arrayExpSliceMinSliceAssign_s", "_arrayExpSliceMinSliceAssign_t", "_arrayExpSliceMinSliceAssign_u", "_arrayExpSliceMinSliceAssign_w", "_arrayExpSliceMinass_a", "_arrayExpSliceMinass_d", // T[]-=T "_arrayExpSliceMinass_f", // T[]-=T "_arrayExpSliceMinass_g", "_arrayExpSliceMinass_h", "_arrayExpSliceMinass_i", "_arrayExpSliceMinass_k", "_arrayExpSliceMinass_s", "_arrayExpSliceMinass_t", "_arrayExpSliceMinass_u", "_arrayExpSliceMinass_w", "_arrayExpSliceMulass_d", // T[]*=T "_arrayExpSliceMulass_f", // T[]*=T "_arrayExpSliceMulass_i", "_arrayExpSliceMulass_k", "_arrayExpSliceMulass_s", "_arrayExpSliceMulass_t", "_arrayExpSliceMulass_u", "_arrayExpSliceMulass_w", "_arraySliceExpAddSliceAssign_a", "_arraySliceExpAddSliceAssign_d", // T[]=T[]+T "_arraySliceExpAddSliceAssign_f", // T[]=T[]+T "_arraySliceExpAddSliceAssign_g", "_arraySliceExpAddSliceAssign_h", "_arraySliceExpAddSliceAssign_i", "_arraySliceExpAddSliceAssign_k", "_arraySliceExpAddSliceAssign_s", "_arraySliceExpAddSliceAssign_t", "_arraySliceExpAddSliceAssign_u", "_arraySliceExpAddSliceAssign_w", "_arraySliceExpDivSliceAssign_d", // T[]=T[]/T "_arraySliceExpDivSliceAssign_f", // T[]=T[]/T "_arraySliceExpMinSliceAssign_a", "_arraySliceExpMinSliceAssign_d", // T[]=T[]-T "_arraySliceExpMinSliceAssign_f", // T[]=T[]-T "_arraySliceExpMinSliceAssign_g", "_arraySliceExpMinSliceAssign_h", "_arraySliceExpMinSliceAssign_i", "_arraySliceExpMinSliceAssign_k", "_arraySliceExpMinSliceAssign_s", "_arraySliceExpMinSliceAssign_t", "_arraySliceExpMinSliceAssign_u", "_arraySliceExpMinSliceAssign_w", "_arraySliceExpMulSliceAddass_d", // T[] += T[]*T "_arraySliceExpMulSliceAddass_f", "_arraySliceExpMulSliceAddass_r", "_arraySliceExpMulSliceAssign_d", // T[]=T[]*T "_arraySliceExpMulSliceAssign_f", // T[]=T[]*T "_arraySliceExpMulSliceAssign_i", "_arraySliceExpMulSliceAssign_k", "_arraySliceExpMulSliceAssign_s", "_arraySliceExpMulSliceAssign_t", "_arraySliceExpMulSliceAssign_u", "_arraySliceExpMulSliceAssign_w", "_arraySliceExpMulSliceMinass_d", // T[] -= T[]*T "_arraySliceExpMulSliceMinass_f", "_arraySliceExpMulSliceMinass_r", "_arraySliceSliceAddSliceAssign_a", "_arraySliceSliceAddSliceAssign_d", // T[]=T[]+T[] "_arraySliceSliceAddSliceAssign_f", // T[]=T[]+T[] "_arraySliceSliceAddSliceAssign_g", "_arraySliceSliceAddSliceAssign_h", "_arraySliceSliceAddSliceAssign_i", "_arraySliceSliceAddSliceAssign_k", "_arraySliceSliceAddSliceAssign_r", // T[]=T[]+T[] "_arraySliceSliceAddSliceAssign_s", "_arraySliceSliceAddSliceAssign_t", "_arraySliceSliceAddSliceAssign_u", "_arraySliceSliceAddSliceAssign_w", "_arraySliceSliceAddass_a", "_arraySliceSliceAddass_d", // T[]+=T[] "_arraySliceSliceAddass_f", // T[]+=T[] "_arraySliceSliceAddass_g", "_arraySliceSliceAddass_h", "_arraySliceSliceAddass_i", "_arraySliceSliceAddass_k", "_arraySliceSliceAddass_s", "_arraySliceSliceAddass_t", "_arraySliceSliceAddass_u", "_arraySliceSliceAddass_w", "_arraySliceSliceMinSliceAssign_a", "_arraySliceSliceMinSliceAssign_d", // T[]=T[]-T[] "_arraySliceSliceMinSliceAssign_f", // T[]=T[]-T[] "_arraySliceSliceMinSliceAssign_g", "_arraySliceSliceMinSliceAssign_h", "_arraySliceSliceMinSliceAssign_i", "_arraySliceSliceMinSliceAssign_k", "_arraySliceSliceMinSliceAssign_r", // T[]=T[]-T[] "_arraySliceSliceMinSliceAssign_s", "_arraySliceSliceMinSliceAssign_t", "_arraySliceSliceMinSliceAssign_u", "_arraySliceSliceMinSliceAssign_w", "_arraySliceSliceMinass_a", "_arraySliceSliceMinass_d", // T[]-=T[] "_arraySliceSliceMinass_f", // T[]-=T[] "_arraySliceSliceMinass_g", "_arraySliceSliceMinass_h", "_arraySliceSliceMinass_i", "_arraySliceSliceMinass_k", "_arraySliceSliceMinass_s", "_arraySliceSliceMinass_t", "_arraySliceSliceMinass_u", "_arraySliceSliceMinass_w", "_arraySliceSliceMulSliceAssign_d", // T[]=T[]*T[] "_arraySliceSliceMulSliceAssign_f", // T[]=T[]*T[] "_arraySliceSliceMulSliceAssign_i", "_arraySliceSliceMulSliceAssign_k", "_arraySliceSliceMulSliceAssign_s", "_arraySliceSliceMulSliceAssign_t", "_arraySliceSliceMulSliceAssign_u", "_arraySliceSliceMulSliceAssign_w", "_arraySliceSliceMulass_d", // T[]*=T[] "_arraySliceSliceMulass_f", // T[]*=T[] "_arraySliceSliceMulass_i", "_arraySliceSliceMulass_k", "_arraySliceSliceMulass_s", "_arraySliceSliceMulass_t", "_arraySliceSliceMulass_u", "_arraySliceSliceMulass_w", }; int i = binary(name, libArrayopFuncs, sizeof(libArrayopFuncs) / sizeof(char *)); if (i == -1) { #ifdef DEBUG // Make sure our array is alphabetized for (i = 0; i < sizeof(libArrayopFuncs) / sizeof(char *); i++) { if (strcmp(name, libArrayopFuncs[i]) == 0) assert(0); } #endif /* Not in library, so generate it. * Construct the function body: * foreach (i; 0 .. p.length) for (size_t i = 0; i < p.length; i++) * loopbody; * return p; */ Arguments *fparams = new Arguments(); Expression *loopbody = buildArrayLoop(fparams); Argument *p = (Argument *)fparams->data[0 /*fparams->dim - 1*/]; #if DMDV1 // for (size_t i = 0; i < p.length; i++) Initializer *init = new ExpInitializer(0, new IntegerExp(0, 0, Type::tsize_t)); Dsymbol *d = new VarDeclaration(0, Type::tsize_t, Id::p, init); Statement *s1 = new ForStatement(0, new DeclarationStatement(0, d), new CmpExp(TOKlt, 0, new IdentifierExp(0, Id::p), new ArrayLengthExp(0, new IdentifierExp(0, p->ident))), new PostExp(TOKplusplus, 0, new IdentifierExp(0, Id::p)), new ExpStatement(0, loopbody)); #else // foreach (i; 0 .. p.length) Statement *s1 = new ForeachRangeStatement(0, TOKforeach, new Argument(0, NULL, Id::p, NULL), new IntegerExp(0, 0, Type::tint32), new ArrayLengthExp(0, new IdentifierExp(0, p->ident)), new ExpStatement(0, loopbody)); #endif Statement *s2 = new ReturnStatement(0, new IdentifierExp(0, p->ident)); //printf("s2: %s\n", s2->toChars()); Statement *fbody = new CompoundStatement(0, s1, s2); /* Construct the function */ TypeFunction *ftype = new TypeFunction(fparams, type, 0, LINKc); //printf("ftype: %s\n", ftype->toChars()); fd = new FuncDeclaration(0, 0, Lexer::idPool(name), STCundefined, ftype); fd->fbody = fbody; fd->protection = PROTpublic; fd->linkage = LINKc; sc->module->importedFrom->members->push(fd); sc = sc->push(); sc->parent = sc->module->importedFrom; sc->stc = 0; sc->linkage = LINKc; fd->semantic(sc); fd->semantic2(sc); fd->semantic3(sc); sc->pop(); } else { /* In library, refer to it. */ fd = FuncDeclaration::genCfunc(type, name); } sv->ptrvalue = fd; // cache symbol in hash table } /* Call the function fd(arguments) */ Expression *ec = new VarExp(0, fd); Expression *e = new CallExp(loc, ec, arguments); e->type = type; return e; }
void VarDeclaration::semantic(Scope *sc) { #if 0 printf("VarDeclaration::semantic('%s', parent = '%s')\n", toChars(), sc->parent->toChars()); printf(" type = %s\n", type ? type->toChars() : "null"); printf(" stc = x%x\n", sc->stc); printf(" storage_class = x%x\n", storage_class); printf("linkage = %d\n", sc->linkage); //if (strcmp(toChars(), "mul") == 0) halt(); #endif storage_class |= sc->stc; if (storage_class & STCextern && init) error("extern symbols cannot have initializers"); AggregateDeclaration *ad = isThis(); if (ad) storage_class |= ad->storage_class & STC_TYPECTOR; /* If auto type inference, do the inference */ int inferred = 0; if (!type) { inuse++; type = init->inferType(sc); inuse--; inferred = 1; /* This is a kludge to support the existing syntax for RAII * declarations. */ storage_class &= ~STCauto; originalType = type; } else { if (!originalType) originalType = type; type = type->semantic(loc, sc); } //printf(" semantic type = %s\n", type ? type->toChars() : "null"); type->checkDeprecated(loc, sc); linkage = sc->linkage; this->parent = sc->parent; //printf("this = %p, parent = %p, '%s'\n", this, parent, parent->toChars()); protection = sc->protection; //printf("sc->stc = %x\n", sc->stc); //printf("storage_class = x%x\n", storage_class); #if DMDV2 if (storage_class & STCgshared && global.params.safe && !sc->module->safe) { error("__gshared not allowed in safe mode; use shared"); } #endif Dsymbol *parent = toParent(); FuncDeclaration *fd = parent->isFuncDeclaration(); Type *tb = type->toBasetype(); if (tb->ty == Tvoid && !(storage_class & STClazy)) { error("voids have no value"); type = Type::terror; tb = type; } if (tb->ty == Tfunction) { error("cannot be declared to be a function"); type = Type::terror; tb = type; } if (tb->ty == Tstruct) { TypeStruct *ts = (TypeStruct *)tb; if (!ts->sym->members) { error("no definition of struct %s", ts->toChars()); } } if ((storage_class & STCauto) && !inferred) error("storage class 'auto' has no effect if type is not inferred, did you mean 'scope'?"); if (tb->ty == Ttuple) { /* Instead, declare variables for each of the tuple elements * and add those. */ TypeTuple *tt = (TypeTuple *)tb; size_t nelems = Parameter::dim(tt->arguments); Objects *exps = new Objects(); exps->setDim(nelems); Expression *ie = init ? init->toExpression() : NULL; for (size_t i = 0; i < nelems; i++) { Parameter *arg = Parameter::getNth(tt->arguments, i); OutBuffer buf; buf.printf("_%s_field_%zu", ident->toChars(), i); buf.writeByte(0); const char *name = (const char *)buf.extractData(); Identifier *id = Lexer::idPool(name); Expression *einit = ie; if (ie && ie->op == TOKtuple) { einit = (Expression *)((TupleExp *)ie)->exps->data[i]; } Initializer *ti = init; if (einit) { ti = new ExpInitializer(einit->loc, einit); } VarDeclaration *v = new VarDeclaration(loc, arg->type, id, ti); //printf("declaring field %s of type %s\n", v->toChars(), v->type->toChars()); v->semantic(sc); #if !IN_LLVM // removed for LDC since TupleDeclaration::toObj already creates the fields; // adding them to the scope again leads to duplicates if (sc->scopesym) { //printf("adding %s to %s\n", v->toChars(), sc->scopesym->toChars()); if (sc->scopesym->members) sc->scopesym->members->push(v); } #endif Expression *e = new DsymbolExp(loc, v); exps->data[i] = e; } TupleDeclaration *v2 = new TupleDeclaration(loc, ident, exps); v2->isexp = 1; aliassym = v2; return; } if (storage_class & STCconst && !init && !fd) // Initialize by constructor only storage_class = (storage_class & ~STCconst) | STCctorinit; if (isConst()) { } else if (isStatic()) { } else if (isSynchronized()) { error("variable %s cannot be synchronized", toChars()); } else if (isOverride()) { error("override cannot be applied to variable"); } else if (isAbstract()) { error("abstract cannot be applied to variable"); } else if (storage_class & STCtemplateparameter) { } else if (storage_class & STCctfe) { } else { AggregateDeclaration *aad = sc->anonAgg; if (!aad) aad = parent->isAggregateDeclaration(); if (aad) { #if DMDV2 assert(!(storage_class & (STCextern | STCstatic | STCtls | STCgshared))); if (storage_class & (STCconst | STCimmutable) && init) { if (!type->toBasetype()->isTypeBasic()) storage_class |= STCstatic; } else #endif aad->addField(sc, this); } InterfaceDeclaration *id = parent->isInterfaceDeclaration(); if (id) { error("field not allowed in interface"); } /* Templates cannot add fields to aggregates */ TemplateInstance *ti = parent->isTemplateInstance(); if (ti) { // Take care of nested templates while (1) { TemplateInstance *ti2 = ti->tempdecl->parent->isTemplateInstance(); if (!ti2) break; ti = ti2; } // If it's a member template AggregateDeclaration *ad = ti->tempdecl->isMember(); if (ad && storage_class != STCundefined) { error("cannot use template to add field to aggregate '%s'", ad->toChars()); } } } #if DMDV2 if ((storage_class & (STCref | STCparameter | STCforeach)) == STCref && ident != Id::This) { error("only parameters or foreach declarations can be ref"); } #endif if (type->isscope() && !noscope) { if (storage_class & (STCfield | STCout | STCref | STCstatic) || !fd) { error("globals, statics, fields, ref and out parameters cannot be auto"); } if (!(storage_class & STCscope)) { if (!(storage_class & STCparameter) && ident != Id::withSym) error("reference to scope class must be scope"); } } enum TOK op = TOKconstruct; if (!init && !sc->inunion && !isStatic() && !isConst() && fd && !(storage_class & (STCfield | STCin | STCforeach)) && type->size() != 0) { // Provide a default initializer //printf("Providing default initializer for '%s'\n", toChars()); if (type->ty == Tstruct && ((TypeStruct *)type)->sym->zeroInit == 1) { /* If a struct is all zeros, as a special case * set it's initializer to the integer 0. * In AssignExp::toElem(), we check for this and issue * a memset() to initialize the struct. * Must do same check in interpreter. */ Expression *e = new IntegerExp(loc, 0, Type::tint32); Expression *e1; e1 = new VarExp(loc, this); e = new AssignExp(loc, e1, e); e->op = TOKconstruct; e->type = e1->type; // don't type check this, it would fail init = new ExpInitializer(loc, e); return; } else if (type->ty == Ttypedef) { TypeTypedef *td = (TypeTypedef *)type; if (td->sym->init) { init = td->sym->init; ExpInitializer *ie = init->isExpInitializer(); if (ie) // Make copy so we can modify it init = new ExpInitializer(ie->loc, ie->exp); } else init = getExpInitializer(); } else { init = getExpInitializer(); } // Default initializer is always a blit op = TOKblit; } if (init) { sc = sc->push(); sc->stc &= ~(STC_TYPECTOR | STCpure | STCnothrow | STCref); ArrayInitializer *ai = init->isArrayInitializer(); if (ai && tb->ty == Taarray) { init = ai->toAssocArrayInitializer(); } StructInitializer *si = init->isStructInitializer(); ExpInitializer *ei = init->isExpInitializer(); // See if initializer is a NewExp that can be allocated on the stack if (ei && isScope() && ei->exp->op == TOKnew) { NewExp *ne = (NewExp *)ei->exp; if (!(ne->newargs && ne->newargs->dim)) { ne->onstack = 1; onstack = 1; if (type->isBaseOf(ne->newtype->semantic(loc, sc), NULL)) onstack = 2; } } // If inside function, there is no semantic3() call if (sc->func) { // If local variable, use AssignExp to handle all the various // possibilities. if (fd && !isStatic() && !isConst() && !init->isVoidInitializer()) { //printf("fd = '%s', var = '%s'\n", fd->toChars(), toChars()); if (!ei) { Expression *e = init->toExpression(); if (!e) { init = init->semantic(sc, type); e = init->toExpression(); if (!e) { error("is not a static and cannot have static initializer"); return; } } ei = new ExpInitializer(init->loc, e); init = ei; } Expression *e1 = new VarExp(loc, this); Type *t = type->toBasetype(); if (t->ty == Tsarray && !(storage_class & (STCref | STCout))) { ei->exp = ei->exp->semantic(sc); if (!ei->exp->implicitConvTo(type)) { int dim = ((TypeSArray *)t)->dim->toInteger(); // If multidimensional static array, treat as one large array while (1) { t = t->nextOf()->toBasetype(); if (t->ty != Tsarray) break; dim *= ((TypeSArray *)t)->dim->toInteger(); e1->type = new TypeSArray(t->nextOf(), new IntegerExp(0, dim, Type::tindex)); } } e1 = new SliceExp(loc, e1, NULL, NULL); } else if (t->ty == Tstruct) { ei->exp = ei->exp->semantic(sc); ei->exp = resolveProperties(sc, ei->exp); StructDeclaration *sd = ((TypeStruct *)t)->sym; #if DMDV2 /* Look to see if initializer is a call to the constructor */ if (sd->ctor && // there are constructors ei->exp->type->ty == Tstruct && // rvalue is the same struct ((TypeStruct *)ei->exp->type)->sym == sd && ei->exp->op == TOKstar) { /* Look for form of constructor call which is: * *__ctmp.ctor(arguments...) */ PtrExp *pe = (PtrExp *)ei->exp; if (pe->e1->op == TOKcall) { CallExp *ce = (CallExp *)pe->e1; if (ce->e1->op == TOKdotvar) { DotVarExp *dve = (DotVarExp *)ce->e1; if (dve->var->isCtorDeclaration()) { /* It's a constructor call, currently constructing * a temporary __ctmp. */ /* Before calling the constructor, initialize * variable with a bit copy of the default * initializer */ Expression *e = new AssignExp(loc, new VarExp(loc, this), t->defaultInit(loc)); e->op = TOKblit; e->type = t; ei->exp = new CommaExp(loc, e, ei->exp); /* Replace __ctmp being constructed with e1 */ dve->e1 = e1; return; } } } } #endif if (!ei->exp->implicitConvTo(type)) { /* Look for opCall * See bugzilla 2702 for more discussion */ Type *ti = ei->exp->type->toBasetype(); // Don't cast away invariant or mutability in initializer if (search_function(sd, Id::call) && /* Initializing with the same type is done differently */ !(ti->ty == Tstruct && t->toDsymbol(sc) == ti->toDsymbol(sc))) { // Rewrite as e1.call(arguments) Expression * eCall = new DotIdExp(loc, e1, Id::call); ei->exp = new CallExp(loc, eCall, ei->exp); } } } ei->exp = new AssignExp(loc, e1, ei->exp); ei->exp->op = TOKconstruct; canassign++; ei->exp = ei->exp->semantic(sc); canassign--; ei->exp->optimize(WANTvalue); } else { init = init->semantic(sc, type); if (fd && isConst() && !isStatic()) { // Make it static storage_class |= STCstatic; } } } else if (isConst() || isFinal() || parent->isAggregateDeclaration()) { /* Because we may need the results of a const declaration in a * subsequent type, such as an array dimension, before semantic2() * gets ordinarily run, try to run semantic2() now. * Ignore failure. */ if (!global.errors && !inferred) { unsigned errors = global.errors; global.gag++; //printf("+gag\n"); Expression *e; Initializer *i2 = init; inuse++; if (ei) { e = ei->exp->syntaxCopy(); e = e->semantic(sc); e = e->implicitCastTo(sc, type); } else if (si || ai) { i2 = init->syntaxCopy(); i2 = i2->semantic(sc, type); } inuse--; global.gag--; //printf("-gag\n"); if (errors != global.errors) // if errors happened { if (global.gag == 0) global.errors = errors; // act as if nothing happened #if DMDV2 /* Save scope for later use, to try again */ scope = new Scope(*sc); scope->setNoFree(); #endif } else if (ei) { if (isDataseg() || (storage_class & STCmanifest)) e = e->optimize(WANTvalue | WANTinterpret); else e = e->optimize(WANTvalue); switch (e->op) { case TOKint64: case TOKfloat64: case TOKstring: case TOKarrayliteral: case TOKassocarrayliteral: case TOKstructliteral: case TOKnull: ei->exp = e; // no errors, keep result break; default: #if DMDV2 /* Save scope for later use, to try again */ scope = new Scope(*sc); scope->setNoFree(); #endif break; } } else init = i2; // no errors, keep result } } sc = sc->pop(); } }
void Module::parse() { //printf("Module::parse()\n"); char *srcname = srcfile->name->toChars(); //printf("Module::parse(srcname = '%s')\n", srcname); utf8_t *buf = (utf8_t *)srcfile->buffer; size_t buflen = srcfile->len; if (buflen >= 2) { /* Convert all non-UTF-8 formats to UTF-8. * BOM : http://www.unicode.org/faq/utf_bom.html * 00 00 FE FF UTF-32BE, big-endian * FF FE 00 00 UTF-32LE, little-endian * FE FF UTF-16BE, big-endian * FF FE UTF-16LE, little-endian * EF BB BF UTF-8 */ unsigned le; unsigned bom = 1; // assume there's a BOM if (buf[0] == 0xFF && buf[1] == 0xFE) { if (buflen >= 4 && buf[2] == 0 && buf[3] == 0) { // UTF-32LE le = 1; Lutf32: OutBuffer dbuf; unsigned *pu = (unsigned *)(buf); unsigned *pumax = &pu[buflen / 4]; if (buflen & 3) { error("odd length of UTF-32 char source %u", buflen); fatal(); } dbuf.reserve(buflen / 4); for (pu += bom; pu < pumax; pu++) { unsigned u; u = le ? readlongLE(pu) : readlongBE(pu); if (u & ~0x7F) { if (u > 0x10FFFF) { error("UTF-32 value %08x greater than 0x10FFFF", u); fatal(); } dbuf.writeUTF8(u); } else dbuf.writeByte(u); } dbuf.writeByte(0); // add 0 as sentinel for scanner buflen = dbuf.offset - 1; // don't include sentinel in count buf = (utf8_t *) dbuf.extractData(); } else { // UTF-16LE (X86) // Convert it to UTF-8 le = 1; Lutf16: OutBuffer dbuf; unsigned short *pu = (unsigned short *)(buf); unsigned short *pumax = &pu[buflen / 2]; if (buflen & 1) { error("odd length of UTF-16 char source %u", buflen); fatal(); } dbuf.reserve(buflen / 2); for (pu += bom; pu < pumax; pu++) { unsigned u; u = le ? readwordLE(pu) : readwordBE(pu); if (u & ~0x7F) { if (u >= 0xD800 && u <= 0xDBFF) { unsigned u2; if (++pu > pumax) { error("surrogate UTF-16 high value %04x at EOF", u); fatal(); } u2 = le ? readwordLE(pu) : readwordBE(pu); if (u2 < 0xDC00 || u2 > 0xDFFF) { error("surrogate UTF-16 low value %04x out of range", u2); fatal(); } u = (u - 0xD7C0) << 10; u |= (u2 - 0xDC00); } else if (u >= 0xDC00 && u <= 0xDFFF) { error("unpaired surrogate UTF-16 value %04x", u); fatal(); } else if (u == 0xFFFE || u == 0xFFFF) { error("illegal UTF-16 value %04x", u); fatal(); } dbuf.writeUTF8(u); } else dbuf.writeByte(u); } dbuf.writeByte(0); // add 0 as sentinel for scanner buflen = dbuf.offset - 1; // don't include sentinel in count buf = (utf8_t *) dbuf.extractData(); } } else if (buf[0] == 0xFE && buf[1] == 0xFF) { // UTF-16BE le = 0; goto Lutf16; } else if (buflen >= 4 && buf[0] == 0 && buf[1] == 0 && buf[2] == 0xFE && buf[3] == 0xFF) { // UTF-32BE le = 0; goto Lutf32; } else if (buflen >= 3 && buf[0] == 0xEF && buf[1] == 0xBB && buf[2] == 0xBF) { // UTF-8 buf += 3; buflen -= 3; } else { /* There is no BOM. Make use of Arcane Jill's insight that * the first char of D source must be ASCII to * figure out the encoding. */ bom = 0; if (buflen >= 4) { if (buf[1] == 0 && buf[2] == 0 && buf[3] == 0) { // UTF-32LE le = 1; goto Lutf32; } else if (buf[0] == 0 && buf[1] == 0 && buf[2] == 0) { // UTF-32BE le = 0; goto Lutf32; } } if (buflen >= 2) { if (buf[1] == 0) { // UTF-16LE le = 1; goto Lutf16; } else if (buf[0] == 0) { // UTF-16BE le = 0; goto Lutf16; } } // It's UTF-8 if (buf[0] >= 0x80) { error("source file must start with BOM or ASCII character, not \\x%02X", buf[0]); fatal(); } } } /* If it starts with the string "Ddoc", then it's a documentation * source file. */ if (buflen >= 4 && memcmp(buf, "Ddoc", 4) == 0) { comment = buf + 4; isDocFile = 1; if (!docfile) setDocfile(); return; } { Parser p(this, buf, buflen, docfile != NULL); p.nextToken(); members = p.parseModule(); md = p.md; numlines = p.scanloc.linnum; } if (srcfile->ref == 0) ::free(srcfile->buffer); srcfile->buffer = NULL; srcfile->len = 0; /* The symbol table into which the module is to be inserted. */ DsymbolTable *dst; if (md) { /* A ModuleDeclaration, md, was provided. * The ModuleDeclaration sets the packages this module appears in, and * the name of this module. */ this->ident = md->id; this->safe = md->safe; Package *ppack = NULL; dst = Package::resolve(md->packages, &this->parent, &ppack); assert(dst); Module *m = ppack ? ppack->isModule() : NULL; if (m && strcmp(m->srcfile->name->name(), "package.d") != 0) { ::error(md->loc, "package name '%s' conflicts with usage as a module name in file %s", ppack->toPrettyChars(), m->srcfile->toChars()); } } else { /* The name of the module is set to the source file name. * There are no packages. */ dst = modules; // and so this module goes into global module symbol table /* Check to see if module name is a valid identifier */ if (!Lexer::isValidIdentifier(this->ident->toChars())) error("has non-identifier characters in filename, use module declaration instead"); } // Insert module into the symbol table Dsymbol *s = this; bool isPackageMod = strcmp(srcfile->name->name(), "package.d") == 0; if (isPackageMod) { /* If the source tree is as follows: * pkg/ * +- package.d * +- common.d * the 'pkg' will be incorporated to the internal package tree in two ways: * import pkg; * and: * import pkg.common; * * If both are used in one compilation, 'pkg' as a module (== pkg/package.d) * and a package name 'pkg' will conflict each other. * * To avoid the conflict: * 1. If preceding package name insertion had occurred by Package::resolve, * later package.d loading will change Package::isPkgMod to PKGmodule and set Package::mod. * 2. Otherwise, 'package.d' wrapped by 'Package' is inserted to the internal tree in here. */ Package *p = new Package(ident); p->parent = this->parent; p->isPkgMod = PKGmodule; p->mod = this; p->symtab = new DsymbolTable(); s = p; } if (!dst->insert(s)) { /* It conflicts with a name that is already in the symbol table. * Figure out what went wrong, and issue error message. */ Dsymbol *prev = dst->lookup(ident); assert(prev); if (Module *mprev = prev->isModule()) { if (strcmp(srcname, mprev->srcfile->toChars()) == 0) error(loc, "from file %s must be imported as module '%s'", srcname, toPrettyChars()); else error(loc, "from file %s conflicts with another module %s from file %s", srcname, mprev->toChars(), mprev->srcfile->toChars()); } else if (Package *pkg = prev->isPackage()) { if (pkg->isPkgMod == PKGunknown && isPackageMod) { /* If the previous inserted Package is not yet determined as package.d, * link it to the actual module. */ pkg->isPkgMod = PKGmodule; pkg->mod = this; } else error(pkg->loc, "from file %s conflicts with package name %s", srcname, pkg->toChars()); } else assert(global.errors); } else { // Add to global array of all modules amodules.push(this); } }
Module *Module::load(Loc loc, Identifiers *packages, Identifier *ident) { Module *m; char *filename; //printf("Module::load(ident = '%s')\n", ident->toChars()); // Build module filename by turning: // foo.bar.baz // into: // foo\bar\baz filename = ident->toChars(); if (packages && packages->dim) { OutBuffer buf; for (size_t i = 0; i < packages->dim; i++) { Identifier *pid = (*packages)[i]; buf.writestring(pid->toChars()); #if _WIN32 buf.writeByte('\\'); #else buf.writeByte('/'); #endif } buf.writestring(filename); buf.writeByte(0); filename = (char *)buf.extractData(); } m = new Module(filename, ident, 0, 0); m->loc = loc; /* Look for the source file */ const char *result = lookForSourceFile(filename); if (result) m->srcfile = new File(result); if (global.params.verbose) { fprintf(global.stdmsg, "import "); if (packages) { for (size_t i = 0; i < packages->dim; i++) { Identifier *pid = (*packages)[i]; fprintf(global.stdmsg, "%s.", pid->toChars()); } } fprintf(global.stdmsg, "%s\t(%s)\n", ident->toChars(), m->srcfile->toChars()); } if (!m->read(loc)) return NULL; m->parse(); #ifdef IN_GCC d_gcc_magic_module(m); #endif return m; }
Expression *StringExp::castTo(Scope *sc, Type *t) { /* This follows copy-on-write; any changes to 'this' * will result in a copy. * The this->string member is considered immutable. */ StringExp *se; Type *tb; int copied = 0; //printf("StringExp::castTo(t = %s), '%s' committed = %d\n", t->toChars(), toChars(), committed); if (!committed && t->ty == Tpointer && t->nextOf()->ty == Tvoid) { error("cannot convert string literal to void*"); return new ErrorExp(); } se = this; if (!committed) { se = (StringExp *)copy(); se->committed = 1; copied = 1; } if (type == t) { return se; } tb = t->toBasetype(); //printf("\ttype = %s\n", type->toChars()); if (tb->ty == Tdelegate && type->toBasetype()->ty != Tdelegate) return Expression::castTo(sc, t); Type *typeb = type->toBasetype(); if (typeb == tb) { if (!copied) { se = (StringExp *)copy(); copied = 1; } se->type = t; return se; } if (tb->ty != Tsarray && tb->ty != Tarray && tb->ty != Tpointer) { if (!copied) { se = (StringExp *)copy(); copied = 1; } goto Lcast; } if (typeb->ty != Tsarray && typeb->ty != Tarray && typeb->ty != Tpointer) { if (!copied) { se = (StringExp *)copy(); copied = 1; } goto Lcast; } if (typeb->nextOf()->size() == tb->nextOf()->size()) { if (!copied) { se = (StringExp *)copy(); copied = 1; } if (tb->ty == Tsarray) goto L2; // handle possible change in static array dimension se->type = t; return se; } if (committed) goto Lcast; #define X(tf,tt) ((tf) * 256 + (tt)) { OutBuffer buffer; size_t newlen = 0; int tfty = typeb->nextOf()->toBasetype()->ty; int ttty = tb->nextOf()->toBasetype()->ty; switch (X(tfty, ttty)) { case X(Tchar, Tchar): case X(Twchar,Twchar): case X(Tdchar,Tdchar): break; case X(Tchar, Twchar): for (size_t u = 0; u < len;) { unsigned c; const char *p = utf_decodeChar((unsigned char *)se->string, len, &u, &c); if (p) error("%s", p); else buffer.writeUTF16(c); } newlen = buffer.offset / 2; buffer.writeUTF16(0); goto L1; case X(Tchar, Tdchar): for (size_t u = 0; u < len;) { unsigned c; const char *p = utf_decodeChar((unsigned char *)se->string, len, &u, &c); if (p) error("%s", p); buffer.write4(c); newlen++; } buffer.write4(0); goto L1; case X(Twchar,Tchar): for (size_t u = 0; u < len;) { unsigned c; const char *p = utf_decodeWchar((unsigned short *)se->string, len, &u, &c); if (p) error("%s", p); else buffer.writeUTF8(c); } newlen = buffer.offset; buffer.writeUTF8(0); goto L1; case X(Twchar,Tdchar): for (size_t u = 0; u < len;) { unsigned c; const char *p = utf_decodeWchar((unsigned short *)se->string, len, &u, &c); if (p) error("%s", p); buffer.write4(c); newlen++; } buffer.write4(0); goto L1; case X(Tdchar,Tchar): for (size_t u = 0; u < len; u++) { unsigned c = ((unsigned *)se->string)[u]; if (!utf_isValidDchar(c)) error("invalid UCS-32 char \\U%08x", c); else buffer.writeUTF8(c); newlen++; } newlen = buffer.offset; buffer.writeUTF8(0); goto L1; case X(Tdchar,Twchar): for (size_t u = 0; u < len; u++) { unsigned c = ((unsigned *)se->string)[u]; if (!utf_isValidDchar(c)) error("invalid UCS-32 char \\U%08x", c); else buffer.writeUTF16(c); newlen++; } newlen = buffer.offset / 2; buffer.writeUTF16(0); goto L1; L1: if (!copied) { se = (StringExp *)copy(); copied = 1; } se->string = buffer.extractData(); se->len = newlen; se->sz = tb->nextOf()->size(); break; default: assert(typeb->nextOf()->size() != tb->nextOf()->size()); goto Lcast; } } #undef X L2: assert(copied); // See if need to truncate or extend the literal if (tb->ty == Tsarray) { int dim2 = ((TypeSArray *)tb)->dim->toInteger(); //printf("dim from = %d, to = %d\n", se->len, dim2); // Changing dimensions if (dim2 != se->len) { // Copy when changing the string literal unsigned newsz = se->sz; void *s; int d; d = (dim2 < se->len) ? dim2 : se->len; s = (unsigned char *)mem.malloc((dim2 + 1) * newsz); memcpy(s, se->string, d * newsz); // Extend with 0, add terminating 0 memset((char *)s + d * newsz, 0, (dim2 + 1 - d) * newsz); se->string = s; se->len = dim2; } } se->type = t; return se; Lcast: Expression *e = new CastExp(loc, se, t); e->type = t; // so semantic() won't be run on e return e; }
void obj_write_deferred(Library *library) { for (size_t i = 0; i < obj_symbols_towrite.dim; i++) { Dsymbol *s = obj_symbols_towrite[i]; Module *m = s->getModule(); char *mname; if (m) { mname = m->srcfile->toChars(); lastmname = mname; } else { //mname = s->ident->toChars(); mname = lastmname; assert(mname); } obj_start(mname); static int count; count++; // sequence for generating names /* Create a module that's a doppelganger of m, with just * enough to be able to create the moduleinfo. */ OutBuffer idbuf; idbuf.printf("%s.%d", m ? m->ident->toChars() : mname, count); char *idstr = idbuf.toChars(); if(!m) { // it doesn't make sense to make up a module if we don't know where to put the symbol // so output it into it's own object file without ModuleInfo objmod->initfile(idstr, NULL, mname); s->toObjFile(0); objmod->termfile(); } else { idbuf.data = NULL; Identifier *id = Identifier::create(idstr, TOKidentifier); Module *md = Module::create(mname, id, 0, 0); md->members = Dsymbols_create(); md->members->push(s); // its only 'member' is s md->doppelganger = 1; // identify this module as doppelganger md->md = m->md; md->aimports.push(m); // it only 'imports' m md->massert = m->massert; md->munittest = m->munittest; md->marray = m->marray; md->genobjfile(0); } /* Set object file name to be source name with sequence number, * as mangled symbol names get way too long. */ const char *fname = FileName::removeExt(mname); OutBuffer namebuf; unsigned hash = 0; for (char *p = s->toChars(); *p; p++) hash += *p; namebuf.printf("%s_%x_%x.%s", fname, count, hash, global.obj_ext); namebuf.writeByte(0); FileName::free((char *)fname); fname = (char *)namebuf.extractData(); //printf("writing '%s'\n", fname); File *objfile = File::create(fname); obj_end(library, objfile); } obj_symbols_towrite.dim = 0; }