Re: [Msys2-users] How to avoid depending on libgcc_s_dw2-1.dll

[Ray D. <min...@gm...>]

On Wed, Nov 5, 2014 at 11:08 PM, Renato Silva <br....@gm...> wrote:
> 2014-10-22 11:48 GMT-02:00 Ray Donnelly <min...@gm...>:
>>
>> Renato's bug report:
>> https://github.com/Alexpux/MINGW-packages/issues/138 I have just
>> closed with a comment explaining why
>
> I’m sorry, but that was an outdated comment. If you had checked my Pidgin++
> package at that time or asked me, you’d notice that there were no
> third-party DLLs anymore. I didn’t understand why no comments were made on
> my originally detailed information, or why my last question here seemed to
> not have raised anything suspicious.

Sorry, I missed the bit where you said there were no external binaries
anymore. I pressed you on that point a few times on IRC before that
and never got the right answer.

For me to look into a bug report, ideally it will be small. If it's
not possible to reduce it (I've never seen this to be the case)
because the problem only happens in a large program then having a
PKGBUILD (with functioning options for debug and !strip) makes it much
easier. The scope of "it fails in pidgin++ on exit" is very large and
it was coupled with a custom build system and externally built
binaries - at least initially. FWIW, I did spend a few days hacking on
making PKGBUILDs for pidgin++ (specifically one for cyrus-sasl) but it
was taking too long so I had to drop it.

This bug (which I was making no assumptions about being the same as
yours) became critical when a lot of i686 packages for both MSYS2 and
MinGW-w64 started to crash. Then isolating a small test-case became
essential and was very easy. For the record the test-case I came up
with is:

int main(void)
{
    HMODULE module = LoadLibraryA("E:/msys32/usr/bin/msys-z.dll");
    return 0;
}

.. which if you had presented me with, I would've definitely took 5
minutes to look into. The customary URL invoked is http://sscce.org/

> However, for the people having similar problems, my bug report was indeed valid and I confirm that GCC 4.9.2-2
> fixed the problem.

Thanks for the confirmation.