void gotDatedbList ( State60 *st ) {
// must only be run on host #0 since we need just one lock table
if ( g_hostdb.m_myHost->m_hostId != 0 ) { char *xx=NULL;*xx=0; }
// load turk lock table if we need to
bool s_init = false;
if ( ! s_init ) {
s_init = true;
if ( ! g_turkLocks.set(8,sizeof(TurkLock),256) )
log("turk: failed to init turk lock table");
if ( ! g_turkLocks.load(g_conf.m_dir,"turkdir/docidlocks.dat"))
log("turk: failed to load turk lock table");
}
time_t now = getTimeGlobal();
// int16_tcut
RdbList *list = &st->m_list;
// the best docid
int64_t best = 0LL;
// scan the list to get urls/docids to turk out
for ( ; ! list->isExhausted() ; ) {
// get rec
char *k = list->getCurrentKey();
// skip that
list->skipCurrentRecord();
// skip if negative
if ( (k[0] & 0x01) == 0x00 ) continue;
// get the docid
int64_t docid = g_datedb.getDocId ( k );
// skip if locked
TurkLock *tt = (TurkLock *)g_turkLock.getValue(&docid);
// if there check time
if ( tt && now - tt->m_lockTime > 3600 ) {
// remove it
g_turkLock.removeKey(&docId);
// nuke tt
tt = NULL;
}
// if still there, skip it and try next one
if ( tt ) continue;
// ok, we got a good docid to dish out
best = docId;
break;
}
SafeBuf sb;
// print description so they can clikc a button to start the turk
sb.safePrintf("<html>\n"
"<title>Event Editor</title>\n"
"<body>\n"
"<table width=\"100%%\" border=\"0\">\n"
"<tr><td style=\"background-color:#0079ba;\">\n"
"<center><font color=#00000>"
"<h2>Event Editor</h2>\n"
"</font></center></td>"
"</tr></table>");
// if we had no docid, give user an empty msg
if ( ! best ) {
sb.safePrintf("<center>Nothing currently available to edit. "
"Please try again later.</center>"
"</body></html>\n");
sendReply ( &sb );
return;
}
// lock it!
TurkLock tt;
strcpy ( tt.m_user , st->m_user );
tt.m_lockTime = now;
if ( ! g_lockTable.addLock ( &tt ) ) {
sendErrorReply ( st , g_errno );
return;
}
// . fetch the TitleRec
// . a max cache age of 0 means not to read from the cache
XmlDoc *xd = &st->m_xd;
// . when getTitleRec() is called it will load the old one
// since XmlDoc::m_setFromTitleRec will be true
// . niceness is 0
xd->set3 ( best , st->m_coll , 0 );
// if it blocks while it loads title rec, it will re-call this routine
xd->setCallback ( st , processLoopWrapper );
// good to go!
return processLoop ( st );
}
void processReply ( char *reply , long replyLen ) {
// store our current reply
SafeBuf fb2;
fb2.safeMemcpy(reply,replyLen );
fb2.nullTerm();
// log that we got the reply
log("qa: got reply(len=%li)(errno=%s)=%s",
replyLen,mstrerror(g_errno),reply);
char *content = NULL;
long contentLen = 0;
// get mime
if ( reply ) {
HttpMime mime;
mime.set ( reply, replyLen , NULL );
// only hash content since mime has a timestamp in it
content = mime.getContent();
contentLen = mime.getContentLen();
if ( content && contentLen>0 && content[contentLen] ) {
char *xx=NULL;*xx=0; }
}
if ( ! content ) {
content = "";
contentLen = 0;
}
s_content = content;
// take out <responseTimeMS>
markOut ( content , "<currentTimeUTC>");
markOut ( content , "<responseTimeMS>");
// until i figure this one out, take it out
markOut ( content , "<docsInCollection>");
// until i figure this one out, take it out
markOut ( content , "<hits>");
// for those links in the html pages
markOut ( content, "rand64=");
// for json
markOut ( content , "\"currentTimeUTC\":" );
markOut ( content , "\"responseTimeMS\":");
markOut ( content , "\"docsInCollection\":");
// for xml
markOut ( content , "<currentTimeUTC>" );
markOut ( content , "<responseTimeMS>");
markOut ( content , "<docsInCollection>");
// indexed 1 day ago
markOut ( content,"indexed:");
// modified 1 day ago
markOut ( content,"modified:");
// s_gigabitCount... it is perpetually incrementing static counter
// in PageResults.cpp
markOut(content,"ccc(");
markOut(content,"id=fd");
markOut(content,"id=sd");
// for some reason the term freq seems to change a little in
// the scoring table
markOut(content,"id=tf");
// make checksum. we ignore back to back spaces so this
// hash works for <docsInCollection>10 vs <docsInCollection>9
long contentCRC = 0;
if ( content ) contentCRC = qa_hash32 ( content );
// note it
log("qa: got contentCRC of %lu",contentCRC);
// if what we expected, save to disk if not there yet, then
// call s_callback() to resume the qa pipeline
/*
if ( contentCRC == s_expectedCRC ) {
// save content if good
char fn3[1024];
sprintf(fn3,"%sqa/content.%lu",g_hostdb.m_dir,contentCRC);
File ff; ff.set ( fn3 );
if ( ! ff.doesExist() ) {
// if not there yet then save it
fb2.save(fn3);
}
// . continue on with the qa process
// . which qa function that may be
//s_callback();
return;
}
*/
//
// if crc of content does not match what was expected then do a diff
// so we can see why not
//
// this means caller does not care about the response
if ( ! s_checkCRC ) {
//s_callback();
return;
}
//const char *emsg = "qa: bad contentCRC of %li should be %li "
// "\n";//"phase=%li\n";
//fprintf(stderr,emsg,contentCRC,s_expectedCRC);//,s_phase-1);
// hash url
long urlHash32 = hash32n ( s_url.getUrl() );
// combine test function too since two tests may use the same url
long nameHash = hash32n ( s_qt->m_testName );
// combine together
urlHash32 = hash32h ( nameHash , urlHash32 );
static bool s_init = false;
if ( ! s_init ) {
s_init = true;
s_ht.set(4,4,1024,NULL,0,false,0,"qaht");
// make symlink
//char cmd[512];
//snprintf(cmd,"cd %s/html ;ln -s ../qa ./qa", g_hostdb.m_dir);
//system(cmd);
char dir[1024];
snprintf(dir,1000,"%sqa",g_hostdb.m_dir);
long status = ::mkdir ( dir ,
S_IRUSR | S_IWUSR | S_IXUSR |
S_IRGRP | S_IWGRP | S_IXGRP |
S_IROTH | S_IXOTH );
if ( status == -1 && errno != EEXIST && errno )
log("qa: Failed to make directory %s: %s.",
dir,mstrerror(errno));
// try to load from disk
SafeBuf fn;
fn.safePrintf("%s/qa/",g_hostdb.m_dir);
log("qa: loading crctable.dat");
s_ht.load ( fn.getBufStart() , "crctable.dat" );
}
// break up into lines
char fn2[1024];
sprintf(fn2,"%sqa/content.%lu",g_hostdb.m_dir,contentCRC);
fb2.save ( fn2 );
// look up in hashtable to see what reply crc should be
long *val = (long *)s_ht.getValue ( &urlHash32 );
// just return if the same
if ( val && contentCRC == *val ) {
g_qaOutput.safePrintf("<b style=color:green;>"
"passed test</b><br>%s : "
"<a href=%s>%s</a> (urlhash=%lu "
"crc=<a href=/qa/content.%lu>"
"%lu</a>)<br>"
"<hr>",
s_qt->m_testName,
s_url.getUrl(),
s_url.getUrl(),
urlHash32,
contentCRC,
contentCRC);
return;
}
if ( ! val ) {
// add it so we know
s_ht.addKey ( &urlHash32 , &contentCRC );
g_qaOutput.safePrintf("<b style=color:blue;>"
"first time testing</b><br>%s : "
"<a href=%s>%s</a> "
"(urlhash=%lu "
"crc=<a href=/qa/content.%lu>%lu"
"</a>)<br>"
"<hr>",
s_qt->m_testName,
s_url.getUrl(),
s_url.getUrl(),
urlHash32,
contentCRC,
contentCRC);
return;
}
log("qa: crc changed for url %s from %li to %li",
s_url.getUrl(),*val,contentCRC);
// get response on file
SafeBuf fb1;
char fn1[1024];
sprintf(fn1,"%sqa/content.%lu",g_hostdb.m_dir, *val);
fb1.load(fn1);
fb1.nullTerm();
// do the diff between the two replies so we can see what changed
char cmd[1024];
sprintf(cmd,"diff %s %s > /tmp/diffout",fn1,fn2);
log("qa: %s\n",cmd);
system(cmd);
g_numErrors++;
g_qaOutput.safePrintf("<b style=color:red;>FAILED TEST</b><br>%s : "
"<a href=%s>%s</a> (urlhash=%lu)<br>"
"<input type=checkbox name=urlhash%lu value=1 "
// use ajax to update test crc. if you undo your
// check then it should put the old val back.
// when you first click the checkbox it should
// gray out the diff i guess.
"onclick=submitchanges(%lu,%lu);> "
"Accept changes"
"<br>"
"original on left, new on right. "
"oldcrc = <a href=/qa/content.%lu>%lu</a>"
" != <a href=/qa/content.%lu>%lu</a> = newcrc"
"<br>diff output follows:<br>"
"<pre id=%lu style=background-color:0xffffff;>",
s_qt->m_testName,
s_url.getUrl(),
s_url.getUrl(),
urlHash32,
// input checkbox name field
urlHash32,
// submitchanges() parms
urlHash32,
contentCRC,
// original/old content.%lu
*val,
*val,
// new content.%lu
contentCRC,
contentCRC,
// for the pre tag id:
urlHash32);
// store in output
SafeBuf sb;
sb.load("/tmp/diffout");
g_qaOutput.htmlEncode ( sb.getBufStart() );
g_qaOutput.safePrintf("</pre><br><hr>");
// if this is zero allow it to slide by. it is learning mode i guess.
// so we can learn what crc we need to use.
// otherwise, stop right there for debugging
//if ( s_expectedCRC != 0 ) exit(1);
// keep on going
//s_callback();
}
