| Summary: | graphviz caller graph only partially drawn (no boxes) | ||
|---|---|---|---|
| Product: | [Developer tools] kcachegrind | Reporter: | rgpublic |
| Component: | general | Assignee: | Josef Weidendorfer <josef.weidendorfer> |
| Status: | RESOLVED FIXED | ||
| Severity: | normal | CC: | flameeyes |
| Priority: | NOR | ||
| Version: | unspecified | ||
| Target Milestone: | --- | ||
| Platform: | Ubuntu | ||
| OS: | Linux | ||
| Latest Commit: | Version Fixed In: | ||
| Sentry Crash Report: | |||
| Attachments: |
Screenshot
Kcachegrind output Generated dot file Dot output |
||
|
Description
rgpublic
2006-03-28 10:36:56 UTC
Created attachment 15334 [details]
Screenshot
Oh. Seems like a graphviz bug. Which version of graphviz is this? Graphviz Version: 2.2.1-1sarge1 This version comes with (K)ubuntu Dapper. What's peculiar: This version also comes with Breezy and works flawlessly there. Created attachment 16454 [details]
Kcachegrind output
This is the cachegrind ouput
i've the same version of kcachegrind/graphviz of rgpublic@gmx.net Same problem here with KCacheGrind 3.5.4 graphviz 2.8 on Gentoo. Hmm... Can someone of you which sees this bug start KCachegrind from a terminal, and when a (buggy) graph is drawn, there should be someting like kcachegrind: Running 'dot /tmp/kde-weidendo/kcachegrind33hR0b.dot -Tplain'... shown in the terminal. The given file exists as long as the graph is shown. Can you please attach (1) the file mentioned as argument to the shown dot command line, and (2) the stdout of this command when rerun in the shell? For smaller data, it would be good to choose a function with a small call graph around. This way, I at least can check if this is a bug in dot/graphviz or not. Thanks, Josef Created attachment 17795 [details]
Generated dot file
Created attachment 17796 [details]
Dot output
Ok ... really strange. On my Suse 10.1, I get exactly the same dot output as Diego (I have dot version 2.6), but my graphs show fine here. So this has nothing to do with dot/graphviz, and yet, I can not reproduce it. Perhaps some other runtime environment issue? What version of GCC, libc, QT are you using? Compiler GCC 4.1.1 (-r1), KDE 3.5.4, Qt 3.3.6 + qtimm patch. if you can point me toward something to look for, might be an unstated dependency missing, or something similar. Exactly the same problem here, using the package from kubuntu edgy (6.10) and also compiling it: dot version 2.8 (Wed Jul 12 21:03:29 UTC 2006) Qt: 3.3.6 KDE: 3.5.5 KCachegrind: 0.4.6 gcc (GCC) 4.1.2 20060928 (prerelease) (Ubuntu 4.1.1-13ubuntu5) The output graph it's ok running the command in the terminal, but not in the program. Any idea? Here's the output with DEBUG_GRAPH = 1 kcachegrind: Loading /home/k7k0/temp/debug/cachegrind.out.firefoxprofiler ... kcachegrind: CallGraphView::refresh kcachegrind: buildGraph(a,0,1) [to Callings] kcachegrind: buildGraph(d,1,1) [to Callings] kcachegrind: buildGraph(a,0,1) [to Callers] kcachegrind: Running 'dot /tmp/kde-k7k0/kcachegrindPRAbma.dot -Tplain'... kcachegrind: Selected a kcachegrind: /tmp/kde-k7k0/kcachegrindPRAbma.dot:1 - graph (0 x 2) => (0 x 140) kcachegrind: /tmp/kde-k7k0/kcachegrindPRAbma.dot:2 - node 'F8303ae8' ( 0/0 - 0x0 ) => (626/552 - 0x0) kcachegrind: /tmp/kde-k7k0/kcachegrindPRAbma.dot:3 - node 'F8303eb8' ( 0/1 - 0x0 ) => (626/482 - 0x0) Here's the callgrind file: version: 0.9.6 cmd: file:/home/k7k0/proyectos/profiler/firefoxprofiler part: 1 events: Time fl=file:/home/k7k0/proyectos/profiler/firefoxprofiler/tests/test.html fn=(null) 3 171 fl=file:/home/k7k0/proyectos/profiler/firefoxprofiler/tests/test.html fn=b 19 6 fl=file:/home/k7k0/proyectos/profiler/firefoxprofiler/tests/test.html fn=c 15 5 fl=file:/home/k7k0/proyectos/profiler/firefoxprofiler/tests/test.html fn=e 5 4 fl=file:/home/k7k0/proyectos/profiler/firefoxprofiler/tests/test.html fn=d 10 25 cfn=e calls=1 0 5 4 fl=file:/home/k7k0/proyectos/profiler/firefoxprofiler/tests/test.html fn=a 23 140 cfn=b calls=1 0 19 6 cfn=c calls=1 0 15 5 cfn=d calls=1 0 10 25 Just fixed that bug. It will be there on KDE 3.5.7 (if exists) or KDE4 otherwise. |