Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20161004173640.GZ19318@brightrain.aerifal.cx>
Date: Tue, 4 Oct 2016 13:36:40 -0400
From: Rich Felker <dalias@...c.org>
To: Julien Ramseier <j.ramseier@...il.com>
Cc: musl@...ts.openwall.com
Subject: Re: Regression: git no longer works with musl libc's regex
 impl

On Tue, Oct 04, 2016 at 07:33:49PM +0200, Julien Ramseier wrote:
> 
> > Le 4 oct. 2016 à 18:08, Johannes Schindelin <Johannes.Schindelin@....de> a écrit :
> > 
> > And lastly, the best alternative would be to teach musl about
> > REG_STARTEND, as it is rather useful a feature.
> 
> I don't remember if I sent a patch for it and was rejected,
> but I patched my musl copy some time ago to support REG_STARTEND.
> The patch is quite small, since the original tre regex implementation
> already supports it.
> 
> Let me know if there is any interest to include REG_STARTEND in musl.

I don't mind looking at it, but I suspect it's fairly costly (maybe
5-10% performance hit) on archs like x86 that are
register-constrained, since you fundamentally need to do extra
spills/reloads to check not only for null termination but also
byte-count limit in the inner loops. This is why I did not keep TRE's
regnexec when I pulled in the code to begin with.

Rich

Powered by blists - more mailing lists

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.