This is the mail archive of the mailing list for the GCC project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: need help finding the patch for Bugzilla 31947/34809/33195

On 03/12/15 13:37, Jonathan Wakely wrote:
On 12 March 2015 at 19:31, JB wrote:
After happily using gcc-4.2 for about 7 years, some recent code changes have
exposed bug 31947 (which has two dups)

I'm investigating the options to address this, and have found that
-fno-inline works in my situation.  An alternate solution would be to patch
the compiler, but I've been unable to identify the change(s) associated with
fixing this problem.  For this particular code base, I'm stuck on 4.2, and
may even have trouble using a patched compiler.   But I didn't expect to get
stuck trying to find the patch.  I've browsed the repo from that time frame
and searched current sources for any mention of any of these bugzilla
numbers to no avail.  Thus my request for help.

When the bug was reported it was only reproducible on the 4.2 branch,
so was already fixed (somehow) on the 4.3 trunk before it was ever
reported. It was never fixed on the 4.2 branch.

That means you're not going to find any mention of the bug in commit
logs, because that would have meant the person making the change that
fixed it for 4.3 would have had to know that PR 31947 would be
reported at some point in the future and guess the PR number!

The brute force approach to finding it would be to do a git bisect
between gcc 4.2.0 RC3 and 4.3 20070515.
And even this procedure produces a nice single commit, the change itself may not be something you're going to want to backport. It may be big or it may depend on other changes or there may have been follow-up fixes.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]