Re: [Pan-users] Weird bug with gcc-5.2's libstdc++ (sort of solved)

[Holger Hoffstätte]

For the archives..

On Thu, 13 Aug 2015 11:01:19 +0000, Holger Hoffstätte wrote:

> On Thu, 13 Aug 2015 10:32:12 +0000, Holger Hoffstätte wrote:
> 
>> Seems like the change in _GLIBCXX_USE_CXX11_ABI deserves a closer look.
> 
> And what do you know..a quick build with -D _GLIBCXX_USE_CXX11_ABI=0
> yields a binary that immediately finds the ~/.pan2 directory regardless
> of $CWD, so there's that.

I've now added -D_GLIBCXX_USE_CXX11_ABI=0 to my local pan package [1] and
it starts up reliably again, irrespective of $CWD.

> Too bad this binary now currupts some prefs file so that the previous
> default-ABI=1 binary now crashes (?).

This turned out to be a fluke: when built against the new ABI pan simply
crashes on startup "sometimes", though it works "most of the time" and
when the $CWD is set to ~.pan2.

This confirms:

> After seeing a bunch of threads start up immediately I'm reasonably
> sure that this gives some credibility to Zan's theory that the previous
> CoW behaviour accidentally saved a temp pointer from getting stomped on.

..which means it's a bug in pan after all.

-h

[1] 
https://github.com/hhoffstaette/portage/blob/master/net-nntp/pan/pan-0.139.ebuild