Re: git cherry(pick) dumps core

From: Andreas Krey <a.krey@gmx.de>
Date: Apr 29, 2010, 20:21 UTC
Message-ID: <20100429202139.GW10879@inner.home.ulmdo.de>
In-Reply-To: <20100429194936.GA31941@progeny.tock>

On Thu, 29 Apr 2010 14:49:36 +0000, Jonathan Nieder wrote:

> Hi Andreas,

...

> Thanks for the report.  What encoding were you using?  (You can check
> with ???git cat-file commit <revision you were trying to cherry-pick>???.)

Actually, it was both UTF-8, both by the defaults in get_message():

reencode_string (
    in=0x14d500 "tree 97cea0f57e06e719c9e7e38c3fc17022048b4f38\nparent 1ef1bb56e95284e683076f0d0cebe26e8ba02eca\nauthor Andreas Krey <a.krey@gmx.de> 1272557698 +0200\ncommitter Andreas Krey <a.krey@gmx.de> 1272557698 +02"..., out_encoding=0x1003f8 "UTF-8", 
        in_encoding=0x1003f8 "UTF-8") at utf8.c:434

I'd guess that in that case, reencode_string wouldn't need to fire up iconv at all, and just copy the original string.

There being two different incarnations of iconv on this machine isn't making anything easier. :-( Will report when I find out what's wrong here. utf8 should be pretty universally known by now.

> In other words, this patch should fix the segfault, but it does not
> fix the underlying problem which was there before.

I was wondering whether to be or not to be, er, to just use the raw message or to die in this case.

Andreas