[LLVMdev] Difficulties Getting a Bug Fix Committed

Renato Golin rengolin at systemcall.org
Tue Dec 4 09:13:32 PST 2012

Hi Michael,

I'm sorry to hear that your patch went unnoticed for such a long time.
I've been through it and I know the pain, but as you said it yourself,
things often fall of the radar and with the number of emails the
commit and the dev lists have, it's not hard to lose track. Often, the
solution to this is to keep bumping and pinging your email until
someone notice it.

About the patch, I'm not sure trial and failure is the best approach
to emitting the exception table. I know GCC does some horrid things
(because of ULEBs and co. changing the final size), but I don't think
this is the problem in hand.

Anton (CCd) is the person that knows most about exception handling, he
might have a better view on what's the best approach to pre-allocate
enough memory for the table (if at all possible).


On 4 December 2012 15:37, Michael Muller <mmuller at enduden.com> wrote:
> Hi all,
> I've been waiting for almost three months now to get a patch committed to LLVM
> that fixes what I consider to be a fairly significant bug in the JIT
> (http://llvm.org/bugs/show_bug.cgi?id=13678) but I've been having a hard time
> getting traction on it.  So far Duncan Sands is the only one who has given it
> any attention, but as of the last time I checked it still wasn't committed and
> the bug is still open.
> Right now we ship our code with a patch to fix this in 3.1, I really don't
> want to have to do that again for 3.2.
> I appreciate the difficulties in running a project as large as LLVM, and it
> looks like there are some recent structural changes that might decentralize
> some of the review/commit burden, but it doesn't speak well for your project
> when you have a bug with a fix and a unit test just ready to go that's not
> getting into the next release.
> Can anyone help me move this forward?
> =============================================================================
> michaelMuller = mmuller at enduden.com | http://www.mindhog.net/~mmuller
> -----------------------------------------------------------------------------
> If you are not willing to control your own mind, there are plenty of other
> people who are willing to do it for you.
> =============================================================================
> _______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev



More information about the llvm-dev mailing list