|
From: robert s. <rbr...@gm...> - 2019-07-10 14:46:51
|
Although I can debug this fine with Totalview, and even use Totalview memory debugger to find detailed location info on memory leaks, Valgrind (memcheck) is only giving me the vaguest info such as (below) : Valgrind-3.13 (actually I compiled 3.15, but it says 3.13 with Valgrind --version) It should be possible to get more detailed info out of Valgrind(memcheck), such as routine and line numbers ?? ==49825== ==49825== 1,066,081 (23,800 direct, 1,042,281 indirect) bytes in 35 blocks are definitely lost in loss record 31,654 of 31,655 ==49825== at 0x4A07155: operator new(unsigned long) (vg_replace_malloc.c:334) ==49825== by 0x1D821C2A: ??? (in /home/rsomervi/GS_code/GS_code_Trunk/hrs/build/lib/linux64/gcc8/hrs/Debug/libseisutil.so) ==49825== by 0x660CE78: ??? (in /home/rsomervi/GS_code/GS_code_Trunk/hrs/build/lib/linux64/gcc8/hrs/Debug/libgseismic_qt4.so) ==49825== by 0x6652CD9: ??? (in /home/rsomervi/GS_code/GS_code_Trunk/hrs/build/lib/linux64/gcc8/hrs/Debug/libgseismic_qt4.so) ==49825== by 0x66A872C: ??? (in /home/rsomervi/GS_code/GS_code_Trunk/hrs/build/lib/linux64/gcc8/hrs/Debug/libgseismic_qt4.so) ==49825== by 0x603F3DF: ??? (in /home/rsomervi/GS_code/GS_code_Trunk/hrs/build/lib/linux64/gcc8/hrs/Debug/libgseismic_qt4.so) ==49825== by 0x657EC7F: ??? (in /home/rsomervi/GS_code/GS_code_Trunk/hrs/build/lib/linux64/gcc8/hrs/Debug/libgseismic_qt4.so) ==49825== by 0x6580C67: ??? (in /home/rsomervi/GS_code/GS_code_Trunk/hrs/build/lib/linux64/gcc8/hrs/Debug/libgseismic_qt4.so) ==49825== by 0x6994D2D: ??? (in /home/rsomervi/GS_code/GS_code_Trunk/hrs/build/lib/linux64/gcc8/hrs/Debug/libgseismic_qt4.so) ==49825== by 0x23587692: QMetaObject::activate(QObject*, int, int, void**) (in /net/GeoSoftware/local/GS_Toolchest/Qt-5.11.3-linux64-gcc8-minimal/lib/libQt5Core.so.5.11.3) ==49825== by 0x6929E4F: ??? (in /home/rsomervi/GS_code/GS_code_Trunk/hrs/build/lib/linux64/gcc8/hrs/Debug/libgseismic_qt4.so) ==49825== by 0x5D0C103: ??? (in /home/rsomervi/GS_code/GS_code_Trunk/hrs/build/lib/linux64/gcc8/hrs/Debug/libgseismic_qt4.so) TIA: Robert Somerville |
|
From: Tom H. <to...@co...> - 2019-07-10 15:13:38
|
On 10/07/2019 15:46, robert somerville wrote: > Although I can debug this fine with Totalview, and even use Totalview > memory debugger to find detailed location info on memory leaks, > Valgrind (memcheck) is only giving me the vaguest info such as (below) : > > Valgrind-3.13 (actually I compiled 3.15, but it says 3.13 with Valgrind > --version) > > It should be possible to get more detailed info out of > Valgrind(memcheck), such as routine and line numbers ?? Well sure, if the files are compiled with debug information. Apparently they aren't, so you don't get anything more. Tom -- Tom Hughes (to...@co...) http://compton.nu/ |
|
From: Julian S. <js...@ac...> - 2019-07-10 15:17:19
|
> Valgrind-3.13 (actually I compiled 3.15, but it says 3.13 with Valgrind > --version) In that case you're not running the version you compiled. Try running 3.15 and give it the flag --keep-debuginfo=yes. Does that help? J |
|
From: Tom H. <to...@co...> - 2019-07-10 15:22:20
|
On 10/07/2019 16:17, Julian Seward wrote: > > > Valgrind-3.13 (actually I compiled 3.15, but it says 3.13 with Valgrind > > --version) > > In that case you're not running the version you compiled. > > Try running 3.15 and give it the flag --keep-debuginfo=yes. Does that > help? If it was an so that had been unloaded then I don't think it would even show the name would it? So I'm not sure --keep-debuginfo=yes will help... Tom -- Tom Hughes (to...@co...) http://compton.nu/ |
|
From: robert s. <rbr...@gm...> - 2019-07-10 15:26:06
|
It actually was compiled with debug info ( as noted, Totalview works fine ). I have tried --keep-debuginfo=yes, but no joy ... I am investigating whether dlclose was called ... On Wed, Jul 10, 2019 at 9:22 AM Tom Hughes <to...@co...> wrote: > On 10/07/2019 16:17, Julian Seward wrote: > > > > > Valgrind-3.13 (actually I compiled 3.15, but it says 3.13 with > Valgrind > > > --version) > > > > In that case you're not running the version you compiled. > > > > Try running 3.15 and give it the flag --keep-debuginfo=yes. Does that > > help? > > If it was an so that had been unloaded then I don't think it would > even show the name would it? So I'm not sure --keep-debuginfo=yes > will help... > > Tom > > -- > Tom Hughes (to...@co...) > http://compton.nu/ > |
|
From: robert s. <rbr...@gm...> - 2019-07-10 21:13:54
|
You were correct! Somehow my environment setup messed up the configure/build/execution(??) of Valgrind 3.15. It now works correctly with --save-debuginfo=yes ! Thanks ;-) On Wed, Jul 10, 2019 at 9:25 AM robert somerville <rbr...@gm...> wrote: > It actually was compiled with debug info ( as noted, Totalview works fine > ). I have tried --keep-debuginfo=yes, but no joy ... > I am investigating whether dlclose was called ... > > On Wed, Jul 10, 2019 at 9:22 AM Tom Hughes <to...@co...> wrote: > >> On 10/07/2019 16:17, Julian Seward wrote: >> > >> > > Valgrind-3.13 (actually I compiled 3.15, but it says 3.13 with >> Valgrind >> > > --version) >> > >> > In that case you're not running the version you compiled. >> > >> > Try running 3.15 and give it the flag --keep-debuginfo=yes. Does that >> > help? >> >> If it was an so that had been unloaded then I don't think it would >> even show the name would it? So I'm not sure --keep-debuginfo=yes >> will help... >> >> Tom >> >> -- >> Tom Hughes (to...@co...) >> http://compton.nu/ >> > |