Path ping...

Jan Hubicka jh@suse.cz
Mon Nov 27 10:06:00 GMT 2006


Roger,
thanks!
> 
> Hmm.  I'm not sure what I think about this.  I suspect that most
> folks regularly add instrumentation code and assertions to gcc's source
> to allow them to investigate a PR or the frequency of operations.
> The question is really at what point does this debugging information
> become useful to other contributors.  It's not uncommon for this
> kind of reporting code to be left attached to a particular PR.
> Is there any policy on this?  I've nothing against the patch which
> looks acceptable, I just can't determine whether much of this data
> would ever be useful to anyone other than you.
> 
> There's also Dirks issues, that the patch doesn't bootstrap as
> submitted, and triggers an undiagnosed problem building libstdc++v3.

My motivation is to have basically similar statistic infrastructure for
all our basic datastructures (we do have same for pools, vectors, ggc
etc.).  I think people use it from time to time and it is very useful
for memory tester, so the changes gets logged quite precisely.  The new
testcase added to GCC triggers a lot of memory consumption in the
bitmaps, so the bitmap related problems popped up few times in recent
weeks/months.

I didnt noticed there was problem with libstc++v3, I will fix that and
send updated patch.

Honza



More information about the Gcc-patches mailing list