vep_do_remove(struct vep_state *vep, enum dowhat what)
{
Debug("DO_REMOVE(%d, end %d empty %d remove %d)\n",
what, vep->endtag, vep->emptytag, vep->remove);
assert(what == DO_TAG);
if (vep->emptytag)
vep_error(vep, "ESI 1.0 <esi:remove/> not legal");
else if (vep->remove && !vep->endtag)
vep_error(vep, "ESI 1.0 <esi:remove> already open");
else if (!vep->remove && vep->endtag)
vep_error(vep, "ESI 1.0 <esi:remove> not open");
else
vep->remove = !vep->endtag;
}
vep_do_comment(struct vep_state *vep, enum dowhat what)
{
Debug("DO_COMMENT(%d)\n", what);
assert(what == DO_TAG);
if (!vep->emptytag)
vep_error(vep, "ESI 1.0 <esi:comment> needs final '/'");
}
struct vsb *
VEP_Finish(struct vep_state *vep)
{
ssize_t l, lcb;
CHECK_OBJ_NOTNULL(vep, VEP_MAGIC);
if (vep->include_src)
VSB_destroy(&vep->include_src);
if (vep->attr_vsb)
VSB_destroy(&vep->attr_vsb);
if (vep->state != VEP_START &&
vep->state != VEP_BOM &&
vep->state != VEP_TESTXML &&
vep->state != VEP_NOTXML &&
vep->state != VEP_NEXTTAG) {
vep_error(vep, "VEP ended inside a tag");
}
if (vep->o_pending)
vep_mark_common(vep, vep->ver_p, vep->last_mark);
if (vep->o_wait > 0) {
lcb = vep->cb(vep->vc, vep->cb_priv, 0, VGZ_ALIGN);
vep_emit_common(vep, lcb - vep->o_last, vep->last_mark);
}
// NB: We don't account for PAD+SUM+LEN in gzip'ed objects
(void)vep->cb(vep->vc, vep->cb_priv, 0, VGZ_FINISH);
AZ(VSB_finish(vep->vsb));
l = VSB_len(vep->vsb);
if (vep->esi_found && l > 0)
return (vep->vsb);
VSB_destroy(&vep->vsb);
return (NULL);
}
void
VEP_Parse(struct vep_state *vep, const char *p, size_t l)
{
const char *e;
struct vep_match *vm;
int i;
CHECK_OBJ_NOTNULL(vep, VEP_MAGIC);
assert(l > 0);
if (vep->startup) {
/*
* We must force the GZIP header out as a SKIP string,
* otherwise an object starting with <esi:include would
* have its GZIP header appear after the included object
* (e000026.vtc)
*/
vep->ver_p = "";
vep->last_mark = SKIP;
vep_mark_common(vep, vep->ver_p, VERBATIM);
vep->startup = 0;
AZ(vep->hack_p);
vep->hack_p = p;
}
vep->ver_p = p;
e = p + l;
while (p < e) {
AN(vep->state);
i = e - p;
if (i > 10)
i = 10;
Debug("EP %s %d (%.*s) [%.*s]\n",
vep->state,
vep->remove,
vep->tag_i, vep->tag,
i, p);
assert(p >= vep->ver_p);
/******************************************************
* SECTION A
*/
if (vep->state == VEP_START) {
if (FEATURE(FEATURE_ESI_REMOVE_BOM) && *p == '\xeb') {
vep->match = vep_match_bom;
vep->state = VEP_MATCH;
} else
vep->state = VEP_BOM;
} else if (vep->state == VEP_BOM) {
vep_mark_skip(vep, p);
if (FEATURE(FEATURE_ESI_DISABLE_XML_CHECK))
vep->state = VEP_NEXTTAG;
else
vep->state = VEP_TESTXML;
} else if (vep->state == VEP_TESTXML) {
/*
* If the first non-whitespace char is different
* from '<' we assume this is not XML.
*/
while (p < e && vct_islws(*p))
p++;
vep_mark_verbatim(vep, p);
if (p < e && *p == '<') {
p++;
vep->state = VEP_STARTTAG;
} else if (p < e && *p == '\xeb') {
VSLb(vep->vc->wrk->vsl, SLT_ESI_xmlerror,
"No ESI processing, "
"first char not '<' but BOM."
" (See feature esi_remove_bom)"
);
vep->state = VEP_NOTXML;
} else if (p < e) {
VSLb(vep->vc->wrk->vsl, SLT_ESI_xmlerror,
"No ESI processing, "
"first char not '<'."
" (See feature esi_disable_xml_check)"
);
vep->state = VEP_NOTXML;
}
} else if (vep->state == VEP_NOTXML) {
/*
* This is not recognized as XML, just skip thru
* vfp_esi_end() will handle the rest
*/
p = e;
vep_mark_verbatim(vep, p);
/******************************************************
* SECTION B
*/
} else if (vep->state == VEP_NOTMYTAG) {
if (FEATURE(FEATURE_ESI_IGNORE_OTHER_ELEMENTS)) {
p++;
vep->state = VEP_NEXTTAG;
} else {
vep->tag_i = 0;
while (p < e) {
if (*p++ == '>') {
vep->state = VEP_NEXTTAG;
break;
}
}
}
if (p == e && !vep->remove)
vep_mark_verbatim(vep, p);
} else if (vep->state == VEP_NEXTTAG) {
/*
* Hunt for start of next tag and keep an eye
* out for end of EsiCmt if armed.
*/
vep->emptytag = 0;
vep->endtag = 0;
vep->attr = NULL;
vep->dostuff = NULL;
while (p < e && *p != '<') {
if (vep->esicmt_p == NULL) {
p++;
continue;
}
if (*p != *vep->esicmt_p) {
p++;
vep->esicmt_p = vep->esicmt;
continue;
}
if (!vep->remove &&
vep->esicmt_p == vep->esicmt)
vep_mark_verbatim(vep, p);
p++;
if (*++vep->esicmt_p == '\0') {
vep->esi_found = 1;
vep->esicmt = NULL;
vep->esicmt_p = NULL;
/*
* The end of the esicmt
* should not be emitted.
* But the stuff before should
*/
vep_mark_skip(vep, p);
}
}
if (p < e) {
if (!vep->remove)
vep_mark_verbatim(vep, p);
assert(*p == '<');
p++;
vep->state = VEP_STARTTAG;
} else if (vep->esicmt_p == vep->esicmt && !vep->remove)
vep_mark_verbatim(vep, p);
/******************************************************
* SECTION C
*/
} else if (vep->state == VEP_STARTTAG) {
/*
* Start of tag, set up match table
*/
if (p < e) {
if (*p == '/') {
vep->endtag = 1;
p++;
}
vep->match = vep_match_starttag;
vep->state = VEP_MATCH;
}
} else if (vep->state == VEP_COMMENT) {
/*
* We are in a comment, find out if it is an
* ESI comment or a regular comment
*/
if (vep->esicmt == NULL)
vep->esicmt_p = vep->esicmt = "esi";
while (p < e) {
if (*p != *vep->esicmt_p) {
vep->esicmt_p = vep->esicmt = NULL;
vep->until_p = vep->until = "-->";
vep->until_s = VEP_NEXTTAG;
vep->state = VEP_UNTIL;
break;
}
p++;
if (*++vep->esicmt_p != '\0')
continue;
if (vep->remove)
vep_error(vep,
"ESI 1.0 Nested <!--esi"
" element in <esi:remove>");
vep->esicmt_p = vep->esicmt = "-->";
vep->state = VEP_NEXTTAG;
vep_mark_skip(vep, p);
break;
}
} else if (vep->state == VEP_CDATA) {
/*
* Easy: just look for the end of CDATA
*/
vep->until_p = vep->until = "]]>";
vep->until_s = VEP_NEXTTAG;
vep->state = VEP_UNTIL;
} else if (vep->state == VEP_ESITAG) {
vep->in_esi_tag = 1;
vep->esi_found = 1;
vep_mark_skip(vep, p);
vep->match = vep_match_esi;
vep->state = VEP_MATCH;
} else if (vep->state == VEP_ESIINCLUDE) {
if (vep->remove) {
vep_error(vep,
"ESI 1.0 <esi:include> element"
" nested in <esi:remove>");
vep->state = VEP_TAGERROR;
} else if (vep->endtag) {
vep_error(vep,
"ESI 1.0 </esi:include> illegal end-tag");
vep->state = VEP_TAGERROR;
} else {
vep->dostuff = vep_do_include;
vep->state = VEP_INTAG;
vep->attr = vep_match_attr_include;
}
} else if (vep->state == VEP_ESIREMOVE) {
vep->dostuff = vep_do_remove;
vep->state = VEP_INTAG;
} else if (vep->state == VEP_ESICOMMENT) {
if (vep->remove) {
vep_error(vep,
"ESI 1.0 <esi:comment> element"
" nested in <esi:remove>");
vep->state = VEP_TAGERROR;
} else if (vep->endtag) {
vep_error(vep,
"ESI 1.0 </esi:comment> illegal end-tag");
vep->state = VEP_TAGERROR;
} else {
vep->dostuff = vep_do_comment;
vep->state = VEP_INTAG;
}
} else if (vep->state == VEP_ESIBOGON) {
vep_error(vep,
"ESI 1.0 <esi:bogus> element");
vep->state = VEP_TAGERROR;
/******************************************************
* SECTION D
*/
} else if (vep->state == VEP_INTAG) {
vep->tag_i = 0;
while (p < e && vct_islws(*p) && !vep->emptytag) {
p++;
vep->canattr = 1;
}
if (p < e && *p == '/' && !vep->emptytag) {
p++;
vep->emptytag = 1;
vep->canattr = 0;
}
if (p < e && *p == '>') {
p++;
AN(vep->dostuff);
vep_mark_skip(vep, p);
vep->dostuff(vep, DO_TAG);
vep->in_esi_tag = 0;
vep->state = VEP_NEXTTAG;
} else if (p < e && vep->emptytag) {
vep_error(vep,
"XML 1.0 '>' does not follow '/' in tag");
vep->state = VEP_TAGERROR;
} else if (p < e && vep->canattr &&
vct_isxmlnamestart(*p)) {
vep->state = VEP_ATTR;
} else if (p < e) {
vep_error(vep,
"XML 1.0 Illegal attribute start char");
vep->state = VEP_TAGERROR;
}
} else if (vep->state == VEP_TAGERROR) {
while (p < e && *p != '>')
p++;
if (p < e) {
p++;
vep_mark_skip(vep, p);
vep->in_esi_tag = 0;
vep->state = VEP_NEXTTAG;
}
/******************************************************
* SECTION E
*/
} else if (vep->state == VEP_ATTR) {
AZ(vep->attr_delim);
if (vep->attr == NULL) {
p++;
AZ(vep->attr_vsb);
vep->state = VEP_SKIPATTR;
} else {
vep->match = vep->attr;
vep->state = VEP_MATCH;
}
} else if (vep->state == VEP_SKIPATTR) {
while (p < e && vct_isxmlname(*p))
p++;
if (p < e && *p == '=') {
p++;
vep->state = VEP_ATTRDELIM;
} else if (p < e && *p == '>') {
vep->state = VEP_INTAG;
} else if (p < e && *p == '/') {
vep->state = VEP_INTAG;
} else if (p < e && vct_issp(*p)) {
vep->state = VEP_INTAG;
} else if (p < e) {
vep_error(vep,
"XML 1.0 Illegal attr char");
vep->state = VEP_TAGERROR;
}
} else if (vep->state == VEP_ATTRGETVAL) {
vep->attr_vsb = VSB_new_auto();
vep->state = VEP_ATTRDELIM;
} else if (vep->state == VEP_ATTRDELIM) {
AZ(vep->attr_delim);
if (*p == '"' || *p == '\'') {
vep->attr_delim = *p++;
vep->state = VEP_ATTRVAL;
} else if (!vct_issp(*p)) {
vep->attr_delim = ' ';
vep->state = VEP_ATTRVAL;
} else {
vep_error(vep,
"XML 1.0 Illegal attribute delimiter");
vep->state = VEP_TAGERROR;
}
} else if (vep->state == VEP_ATTRVAL) {
while (p < e && *p != '>' && *p != vep->attr_delim &&
(vep->attr_delim != ' ' || !vct_issp(*p))) {
if (vep->attr_vsb != NULL)
VSB_bcat(vep->attr_vsb, p, 1);
p++;
}
if (p < e && *p == '>') {
vep_error(vep,
"XML 1.0 Missing end attribute delimiter");
vep->state = VEP_TAGERROR;
vep->attr_delim = 0;
if (vep->attr_vsb != NULL) {
AZ(VSB_finish(vep->attr_vsb));
VSB_delete(vep->attr_vsb);
vep->attr_vsb = NULL;
}
} else if (p < e) {
vep->attr_delim = 0;
p++;
vep->state = VEP_INTAG;
if (vep->attr_vsb != NULL) {
AZ(VSB_finish(vep->attr_vsb));
AN(vep->dostuff);
vep->dostuff(vep, DO_ATTR);
vep->attr_vsb = NULL;
}
}
/******************************************************
* Utility Section
*/
} else if (vep->state == VEP_MATCH) {
/*
* Match against a table
*/
vm = vep_match(vep, p, e);
vep->match_hit = vm;
if (vm != NULL) {
if (vm->match != NULL)
p += strlen(vm->match);
vep->state = *vm->state;
vep->match = NULL;
vep->tag_i = 0;
} else {
memcpy(vep->tag, p, e - p);
vep->tag_i = e - p;
vep->state = VEP_MATCHBUF;
p = e;
}
} else if (vep->state == VEP_MATCHBUF) {
/*
* Match against a table while split over input
* sections.
*/
AN(vep->match);
do {
if (*p == '>') {
for (vm = vep->match;
vm->match != NULL; vm++)
continue;
AZ(vm->match);
} else {
vep->tag[vep->tag_i++] = *p++;
vm = vep_match(vep,
vep->tag, vep->tag + vep->tag_i);
if (vm && vm->match == NULL) {
vep->tag_i--;
p--;
}
}
} while (vm == NULL && p < e);
vep->match_hit = vm;
if (vm == NULL) {
assert(p == e);
} else {
vep->state = *vm->state;
vep->match = NULL;
}
} else if (vep->state == VEP_UNTIL) {
/*
* Skip until we see magic string
*/
while (p < e) {
if (*p++ != *vep->until_p++) {
vep->until_p = vep->until;
} else if (*vep->until_p == '\0') {
vep->state = vep->until_s;
break;
}
}
if (p == e && !vep->remove)
vep_mark_verbatim(vep, p);
} else {
Debug("*** Unknown state %s\n", vep->state);
INCOMPL();
}
}
/*
* We must always mark up the storage we got, try to do so
* in the most efficient way, in particular with respect to
* minimizing and limiting use of pending.
*/
if (p == vep->ver_p)
;
else if (vep->in_esi_tag)
vep_mark_skip(vep, p);
else if (vep->remove)
vep_mark_skip(vep, p);
else
vep_mark_pending(vep, p);
}
vep_do_include(struct vep_state *vep, enum dowhat what)
{
const char *p, *q, *h;
ssize_t l;
Debug("DO_INCLUDE(%d)\n", what);
if (what == DO_ATTR) {
Debug("ATTR (%s) (%s)\n", vep->match_hit->match,
VSB_data(vep->attr_vsb));
if (vep->include_src != NULL) {
vep_error(vep,
"ESI 1.0 <esi:include> "
"has multiple src= attributes");
vep->state = VEP_TAGERROR;
VSB_delete(vep->attr_vsb);
VSB_delete(vep->include_src);
vep->attr_vsb = NULL;
vep->include_src = NULL;
return;
}
vep->include_src = vep->attr_vsb;
return;
}
assert(what == DO_TAG);
if (!vep->emptytag)
vep_warn(vep,
"ESI 1.0 <esi:include> lacks final '/'");
if (vep->include_src == NULL) {
vep_error(vep,
"ESI 1.0 <esi:include> lacks src attr");
return;
}
/*
* Strictly speaking, we ought to spit out any piled up skip before
* emitting the VEC for the include, but objectively that makes no
* difference and robs us of a chance to collapse another skip into
* this on so we don't do that.
* However, we cannot tolerate any verbatim stuff piling up.
* The mark_skip() before calling dostuff should have taken
* care of that. Make sure.
*/
assert(vep->o_wait == 0 || vep->last_mark == SKIP);
/* XXX: what if it contains NUL bytes ?? */
p = VSB_data(vep->include_src);
l = VSB_len(vep->include_src);
h = 0;
if (l > 7 && !memcmp(p, "http://", 7)) {
h = p + 7;
p = strchr(h, '/');
AN(p);
Debug("HOST <%.*s> PATH <%s>\n", (int)(p-h),h, p);
VSB_printf(vep->vsb, "%c", VEC_INCL);
VSB_printf(vep->vsb, "Host: %.*s%c", (int)(p-h), h, 0);
} else if (l > 8 && !memcmp(p, "https://", 8)) {
if (!FEATURE(FEATURE_ESI_IGNORE_HTTPS)) {
vep_warn(vep,
"ESI 1.0 <esi:include> with https:// ignored");
vep->state = VEP_TAGERROR;
vep->attr_vsb = NULL;
vep->include_src = NULL;
return;
}
vep_warn(vep,
"ESI 1.0 <esi:include> https:// treated as http://");
h = p + 8;
p = strchr(h, '/');
AN(p);
VSB_printf(vep->vsb, "%c", VEC_INCL);
VSB_printf(vep->vsb, "Host: %.*s%c", (int)(p-h), h, 0);
} else if (*p == '/') {
VSB_printf(vep->vsb, "%c", VEC_INCL);
VSB_printf(vep->vsb, "%c", 0);
} else {
VSB_printf(vep->vsb, "%c", VEC_INCL);
VSB_printf(vep->vsb, "%c", 0);
/* Look for the last / before a '?' */
h = NULL;
for (q = vep->url; *q && *q != '?'; q++)
if (*q == '/')
h = q;
if (h == NULL)
h = q + 1;
Debug("INCL:: [%.*s]/[%s]\n",
(int)(h - vep->url), vep->url, p);
VSB_printf(vep->vsb, "%.*s/", (int)(h - vep->url), vep->url);
}
l -= (p - VSB_data(vep->include_src));
for (q = p; *q != '\0'; ) {
if (*q == '&') {
#define R(w,f,r) \
if (q + w <= p + l && !memcmp(q, f, w)) { \
VSB_printf(vep->vsb, "%c", r); \
q += w; \
continue; \
}
R(6, "'", '\'');
R(6, """, '"');
R(4, "<", '<');
R(4, ">", '>');
R(5, "&", '&');
}
VSB_printf(vep->vsb, "%c", *q++);
}
#undef R
VSB_printf(vep->vsb, "%c", 0);
VSB_delete(vep->include_src);
vep->include_src = NULL;
}
