Re: segfault in git-remote-http

From: rh <richard_hubbe11@lavabit.com>
Date: Apr 9, 2013, 15:47 UTC
Message-ID: <20130409084718.587e99aa7a935296867a84a1@lavabit.com>
In-Reply-To: <20130407093812.cae0e19123f7b6d2061800aa@lavabit.com>

On Sun, 7 Apr 2013 09:38:12 -0700 rh <richard_hubbe11@lavabit.com> wrote:

Show 21 quoted lines

> Hello,
> (cross-posting to bitcoin-devel in case they've seen it)
>  
> 1) git clone https://github.com/bitcoin/bitcoin.git
> segfaulted in libcrypto.so.1.0.0
> 
> 2) git clone git://github.com/bitcoin/bitcoin.git
> had no problems
> 
> This was yesterday, I retried today
> 
> 3) git clone https://github.com/bitcoin/bitcoin.git
> and it worked fine.
> 
> Yesterday I was running bitcoin-0.8.1-linux/bitcoin-qt when #1 failed.
> Today I was running bitcoind when I ran #3
> 
> Two things I thought might be coming into play here, network
> pressure and memory pressure. Is this occurence reasonably
> possible or is something else afoot?
>

The symptoms that this patch addresses look similar:

http://article.gmane.org/gmane.mail.postfix.user/217790

Quote from that thread: "This behavior is actually documented (SSL_set_fd() destroys a BIO already on the SSL handle, and creates a new BIO)."

Maybe someone used to looking at git-remote-http code can say anything about this.