Re: [Monotone-devel] ding dong, boost::filesystem is dead

[Zack Weinberg]

On 7/12/07, Richard Levitte <address@hidden> wrote:
> In message <address@hidden> on Thu, 12 Jul 2007 14:15:58 -0700, "Zack Weinberg" 
> <address@hidden> said:
>
> zackw> I am entirely in agreement with you. :-)  In fact, the
> zackw> nvm.experiment.pcre branch does just that - replaces
> zackw> boost::regex with libpcre, which is much smaller and can sanely
> zackw> be bundled with monotone.  This is why I am suggesting landing
> zackw> both the .deregexp and the .pcre branches.
>
> Mmmm, I like the idea...  Now, to choose, wait with that until 0.36 is
> released or risk that the build fails miserably?  I frankly dunno what
> to recommend here...  On the other hand, more or less all current
> build slaves, except most the Linux ones, are failing...
>
> You know, I think we might as well go for it now, if you think it's
> safe enough.  Considering there are some build slaves I want to see up
> before doing a release, it will be another week before we get there.

Hm.  The .deregexp branch should be entirely safe, though I'd be
happier with another pair of eyes on it...  However, Nathaniel was
worried about stack-overflow problems with PCRE's use of recursion,
mainly in the context of netsync branch patterns.  Thus, I'd like to
take a crack at rewriting globish_matcher without regular expressions
first.  I can do that tomorrow (maybe) or Monday (more likely).

zw