Colin Burgess(deleted)
|
Re: Mudflap ready for use?
|
Colin Burgess(deleted)
01/23/2009 3:45 PM
post20753
|
Re: Mudflap ready for use?
In my limited experience you are going to get a lot of false positives unless absolutely EVERYTHING is
compiled with -fmudflap
There are somethings (syspage etc) which need to be added to the auto-register portion in the libmudflap routines,
which haven't been yet.
Even when you have registered, btw, it appears that if you fall off the end of an memory object you don't get
told you fell off the end, rather you get told you accessed something that isn't valid, and btw here are some
nearby objects.
Colin
Thomas Fletcher wrote:
> Just wondering if the mudflap support in the gcc provided with SDP 6.4.0 is ready for primetime and/or if there are
any special caveates to be aware of.
>
> I'd like to use it in a multi-threaded app (the mudflapth) variant, but my few experiments with some simple things
seem to be showing some false violations already. Unfortunately, they aren't small enough to show, but I can guarantee
that they aren't violations.
>
> Thanks,
> Thomas
>
> _______________________________________________
> General
> http://community.qnx.com/sf/go/post20752
>
--
cburgess@qnx.com
|
|
|
Elena Laskavaia
|
Re: Mudflap ready for use?
|
Elena Laskavaia
01/23/2009 4:09 PM
post20756
|
Re: Mudflap ready for use?
Mudflap feature released with 6.4 was missing backtrace printing which makes it nearly useless, it was fixed since then
and patch would be available on foundry.
Besides false positives it will generate lots and lots of duplicates, so you need a tool to look at this stuff, which we
going to do in IDE 4.6 (SDP 6.4.1).
It has same false positives but they grouped by backtrace so instead of thousands you looking into 10-20 potential
violations.
In the meantime you can always you QNX Memory Analysis tool :)
Thomas Fletcher wrote:
> Just wondering if the mudflap support in the gcc provided with SDP 6.4.0 is ready for primetime and/or if there are
any special caveates to be aware of.
>
> I'd like to use it in a multi-threaded app (the mudflapth) variant, but my few experiments with some simple things
seem to be showing some false violations already. Unfortunately, they aren't small enough to show, but I can guarantee
that they aren't violations.
>
> Thanks,
> Thomas
>
> _______________________________________________
> General
> http://community.qnx.com/sf/go/post20752
>
|
|
|