|
From: <Mar...@pr...> - 2007-02-15 16:14:28
|
Hello, we were using valgrind as one of the most valuable tools when it comes to=20 debug "mean" errors of our Qt 3.3.2 application. With valgrind 2.4.0 there = was some reasonable amount of noise and it found many errors for us. But with valgrind 3.2.1 (most possibly earlier versions, too) we get so=20 many qt errors that we can't use it. I tried to create a large suppression = file with=20 valgrind --gen-suppressions=3Dall --log-fd=3D4 program.bin 4> supp.tmp egrep "^[^=3D-]" supp.tmp > supp and abort()ed my program just before I suspected an error. Then I started=20 valgrind like this: valgrind --suppressions=3Dsupp program.bin=20 But as soon as the program got past the previously abort()ed section,=20 valgrind again reported loads and loads of errors. So version 3.2.1 seems=20 unusable to us... Do you have any explanation for this ? Has valgrind become so exact or so=20 advanced that it detects so much more errors ? Should we just stick to=20 version 2.4.0 or does it have serious disatvantages ?=20 Many thanks for your help, Markus Grunwald Softwaredevelopment PR=DCFTECHNIK Condition Monitoring GmbH Oskar-Messter-Stra=DFe 19-21 85737 Ismaning www.pruftechnik.com Tel: +49 (0)89 99616177 Fax: +49 (0)89 99616200 |
|
From: Julian S. <js...@ac...> - 2007-02-15 19:18:57
|
> Do you have any explanation for this ? I presume you really mean Valgrind's Memcheck tool. Our policy is to make Memcheck have zero false errors, and generally we are successful with this. If you are getting a lot of errors you do not expect, this may be a bug in Memcheck. I suggest you first try 3.2.3 - this has some fixes that are not in 3.2.1 - and if you still have unexpected errors then you will need to send some details so we can figure out what the problem is. J |
|
From: Nicholas N. <nj...@cs...> - 2007-02-16 00:19:31
|
On Thu, 15 Feb 2007, Julian Seward wrote: > Our policy is to make Memcheck have zero false errors, and generally > we are successful with this. If you are getting a lot of errors you > do not expect, this may be a bug in Memcheck. I suggest you first > try 3.2.3 - this has some fixes that are not in 3.2.1 - and if you still > have unexpected errors then you will need to send some details so > we can figure out what the problem is. And the suppressions should have worked, but we need more information to know what went wrong. Nick |
|
From: <led...@vt...> - 2007-02-16 03:59:24
|
I have been using valgrind 3.2.2 on a RHEL powerpc64 architecture, and I get numerous "conditional or jump depends on uninitialize value" errors. I have tracked the errors down by disassembling the code and a vast majority are false positives. The errors come not only from my code, but from the standard math library (libm). Quoting Mar...@pr...: > Hello, > > we were using valgrind as one of the most valuable tools when it comes to > debug "mean" errors of our Qt 3.3.2 application. With valgrind 2.4.0 there > was some reasonable amount of noise and it found many errors for us. > > But with valgrind 3.2.1 (most possibly earlier versions, too) we get so > many qt errors that we can't use it. I tried to create a large suppression > file with > > valgrind --gen-suppressions=all --log-fd=4 program.bin 4> supp.tmp > egrep "^[^=-]" supp.tmp > supp > > and abort()ed my program just before I suspected an error. Then I started > valgrind like this: > > valgrind --suppressions=supp program.bin > > But as soon as the program got past the previously abort()ed section, > valgrind again reported loads and loads of errors. So version 3.2.1 seems > unusable to us... > > Do you have any explanation for this ? Has valgrind become so exact or so > advanced that it detects so much more errors ? Should we just stick to > version 2.4.0 or does it have serious disatvantages ? > > Many thanks for your help, > > Markus Grunwald > Softwaredevelopment > > PRÜFTECHNIK Condition Monitoring GmbH > Oskar-Messter-Straße 19-21 > 85737 Ismaning > www.pruftechnik.com > Tel: +49 (0)89 99616177 > Fax: +49 (0)89 99616200 > > > ------------------------------------------------------------------------- > Take Surveys. Earn Cash. Influence the Future of IT > Join SourceForge.net's Techsay panel and you'll get the chance to share your > opinions on IT & business topics through brief surveys-and earn cash > http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV > _______________________________________________ > Valgrind-users mailing list > Val...@li... > https://lists.sourceforge.net/lists/listinfo/valgrind-users > |
|
From: Julian S. <js...@ac...> - 2007-02-16 08:27:04
|
> On Friday 16 February 2007 03:59, led...@vt... wrote: > I have been using valgrind 3.2.2 on a RHEL powerpc64 architecture, and I > get numerous "conditional or jump depends on uninitialize value" errors. I > have tracked the errors down by disassembling the code and a vast majority > are false positives. The errors come not only from my code, but from the > standard math library (libm). Yes. I'm not surprised to hear that. Over the years I've tracked down and fixed various causes of false positives, but every time, it requires having a way to reproduce the problem, since figuring out what's going on depends on looking at details of assembly code. J |