This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [commit 7.8] [patch+7.8] Fix 7.8 regression: resolve_dynamic_struct: Assertion `TYPE_NFIELDS (type) > 0' (PR 17642)
- From: Joel Brobecker <brobecker at adacore dot com>
- To: "H.J. Lu" <hjl dot tools at gmail dot com>
- Cc: Jan Kratochvil <jan dot kratochvil at redhat dot com>, GDB <gdb-patches at sourceware dot org>
- Date: Mon, 15 Dec 2014 15:41:43 -0500
- Subject: Re: [commit 7.8] [patch+7.8] Fix 7.8 regression: resolve_dynamic_struct: Assertion `TYPE_NFIELDS (type) > 0' (PR 17642)
- Authentication-results: sourceware.org; auth=none
- References: <20141125195444 dot GA3400 at host2 dot jankratochvil dot net> <20141213142351 dot GG5457 at adacore dot com> <20141215150609 dot GA4229 at adacore dot com> <20141215191244 dot GA23577 at host2 dot jankratochvil dot net> <CAMe9rOod8+arBG7T7FgoNpJmF5LhZ5kwqGfLDF-Wx6ESGH3Qog at mail dot gmail dot com> <20141215193658 dot GC5457 at adacore dot com> <CAMe9rOqefYPoN-qGpH-z2kiHpyB5q7_52YA1ibTTAGX4qeF2Qg at mail dot gmail dot com>
> > This is a fairly big patch...
> > 1. How bad is performance without it?
>
> It was very bad, like > 100x slower.
OK, but 100x slower than 1usec is still unnoticeable. When did you start
noticing it, for what kind of program (size?), and what type of delay
were you seeing? Also, is that a regression compared to 7.7? The idea
is that, if it's just a second or two, or even ten, that's still quite
bearable, and unless you are absolutely sure that your patch is safe,
perhaps we should pass...
> > 2. Can you find a binutils maintainer that will vouch that
> > this patch is 100% safe?
>
> As the x86 binutils maintainer, do I count?
Absolutely. I just need someone who knows the code well enough to
be trusted with its maintainance to stand behind the code and
confirm that it is considered safe. In this case, I need extra safe,
considering the fact that this is a .2 release.
I hope I've described my concerns well enough for you to make
the decision. Now that I have explained my thought process, I feel
you're more qualified to make an informed decision now. Your call.
Please remember that, if you do push it to 7.8, you'll need a PR
associated to is so you can document the fix in the release wiki
page (see fixes in GDB 7.8.2):
https://sourceware.org/gdb/wiki/GDB_7.8_Release
--
Joel