Re: [Valgrind-users] Error limits with CVS head

[Leonard m. <spa...@ya...>]

--- Nicholas Nethercote <nj...@ca...> wrote:
> > Just ran CVS head on an app I have and while the valgrind 
> > output is only 926 lines long (133 flagged issues), 
> > valgrind says:
> >
> >    More than 30000 total errors detected.
> 
> Do you get that many bugs on just one program, or across a range? 
> If it's the former, you've probably just a got a really buggy
> program.  If it's the latter, it's probably a problem with 
> Valgrind itself.

This was running one program.  It's not buggy but it is
large.  And like I said, the file only lists 133 flagged errors,
so the rest of the 30,000 are apparently being suppresssed by
valgrind's default suppression rules.

And of those 133, many of them are things which aren't in this 
program's code, things like:

Conditional jump or move depends on uninitialised value down in ld.so
or the NVidia libGL*, or
Syscall param ioctl(generic) contains uninitialised or unaddressable
byte(s) down in libc with only libc on the stack.

Really, from a usage perspective, it doesn't matter how many issues
valgrind detected if it doesn't report any of them because they're
all suppressed and thus presumably not errors.  I don't think these
should reduce the number of unsuppressed potential errors that the user
does get to see.  But I can turn the limits off so that's OK.
Just a suggestion to make the limits more useful.

Randall





__________________________________
Do you Yahoo!?
Yahoo! Calendar - Free online calendar with sync to Outlook(TM).
http://calendar.yahoo.com