Re: Bug 9954 - getaddrinfo assertion triggered without reason

[Allan McRae <allan at archlinux dot org>]

On 02/05/13 00:12, Carlos O'Donell wrote:
> On 05/01/2013 09:00 AM, Allan McRae wrote:
>> I have a couple of Arch Linux users reporting the same assertion as in
>> this Bug 9954 [1], so I was looking into the patch provided (from 2009...).
>>
>> I appears that both Fedora and Debian carry this patch, although not
>> openSUSE or Gentoo.  A quick search only found this being reported in
>> Fedora, Debian, Ubuntu and Arch (some of those assumably pre-patching).
>>
>> However, I can not replicate using the test case attached to that bug.
>> Is anyone else able to?
> 
> It looks like Fedora has it but RHEL doesn't.
> 
> This is one of these mystery patches that Fedora is carrying that we
> eventually want to merge upstream.
> 
> It looks like reproducing this depends heavily on /etc/hosts having
> certain things in it.
> 
> Perhaps you can use the Fedora bug to guide you in reproducing?
> 
> https://bugzilla.redhat.com/show_bug.cgi?id=739743

I tried to replicate via that but was unable.  I did see a post
suggesting that DNS was involved to replicate, so perhaps I am just
using a good DNS server.

However, looking at the explanation in the bug report, I think this fix
is good whether or not I can replicate...  Can anyone else replicate?

> Let me rant a bit and say:
> 
> I wish all distros had a distro branch in upstream git with all of
> their patches in said git branch and we all worked quickly and easily
> via git pull requests.
> 
> I'll probably set this up for Fedora again and start doing work
> upstream in a Fedora branch and pulling from that into the actual
> Fedora rawhide release, and auto-generating patches from commits.

I can do this for Arch, but it will be limited to a "these are the
patches we think should be backported" branch as we have a policy about
patches requiring acceptance upstream before we use them.

Allan