Yep, tidy is not very tidy. It's been a good source of crashes for as long as I've been aware of it.
Anyone running MarkLogic on OSX might like to know how to simulate pstack on OSX with a simple script. This requires gdb, of course. The output can be very useful for reporting bugs, and sometimes for diagnostics. Just save it as /usr/bin/pstack and make it executable. #!/usr/bin/env bash # # Fake pstack on OSX # http://stackoverflow.com/a/12394981/908390 # ###################################################################### set -e tmp=$(mktemp -t pstack) echo thread apply all bt >"$tmp" gdb -batch -nx -nw -q -x "${tmp}" -p "$1" rm -f "$tmp" # end pstack For this test case it looks like the crash is in strcmp, called by parseStartTag. Thread 80 (process 73728): #0 0x00007fff90131e22 in __wait4 () #1 0x00000001023f65dc in xdmp::pstack () #2 0x00000001023f70b6 in pstackSignalAction () #3 <signal handler called> #4 0x00007fff8d453d13 in _platform_strcmp () #5 0x0000000101e80408 in xdmp::XMLParser::parseStartTag () #6 0x0000000101e7be7a in xdmp::XMLParser::parse () #7 0x0000000101e7c137 in xdmp::parseXML () #8 0x000000010272a644 in xdmp::doTidy () #9 0x000000010272eaa0 in xdmp::xdmp_tidy () -- Mike On 4 Jun 2014, at 15:40 , Jakob Fix <[email protected]> wrote: > try for yourself (7.0-2.3 on Mac): > > let $u := > "http://www.larepublica.co/ocde-recomienda-fortalecer-el-sistema-de-planeaci%C3%B3n_124391"; > > let $c := xdmp:http-get($u, > <options xmlns="xdmp:http" xmlns:d="xdmp:document-get"> > <verify-cert>false</verify-cert> > <d:encoding>UTF-8</d:encoding> > <d:repair>full</d:repair> > </options> > ) > > let $d := xdmp:tidy($c[2], > <options xmlns="xdmp:tidy"> > <new-blocklevel-tags>section, header, time, figure, nav, > article</new-blocklevel-tags> > <bare>yes</bare> > <clean>yes</clean> > <hide-comments>yes</hide-comments> > </options> > )[2] > > return $d > > > log set to "finest" returns just this: > > Segmentation fault in thread 7411003392 addr 0x2aef14847 > /usr/bin/pstack: No such file or directory > 2014-06-05 00:33:07.647 Notice: Starting MarkLogic Server 7.0-2.3 x86_64 in > /Users/jakob/Library/MarkLogic with data in /Users/jakob/Library/Application > Support/MarkLogic/Data > 2014-06-05 00:33:07.655 Info: Host jfix.local running Darwin 13.2.0 > > of course (and most fortunately) this requires some weirdness in the document > passed to xdmp:tidy, but still a pain if you can't control the input you're > getting... > > cheers, > Jakob. > _______________________________________________ > General mailing list > [email protected] > http://developer.marklogic.com/mailman/listinfo/general _______________________________________________ General mailing list [email protected] http://developer.marklogic.com/mailman/listinfo/general
