0.9.5 release plans

[Giorgos Keramidas]

On 2007-10-16 14:48, Mark Reinhold <mr at sun.com> wrote:
>> Date: Mon, 15 Oct 2007 17:31:02 -0500
>> From: Matt Mackall <mpm at selenic.com>
>
>> I'm hoping to finally cut the 0.9.5 release on Friday. We've
>> hopefully fixed up the compilation problems for the new C extensions
>> reported a couple days ago. Please test the current tip, especially
>> if you're running on one of the less common platforms.
>
> main tip (3b204881f959) builds and tests fine on Solaris 10 using
> Python 2.5.1 from blastwave.org, except for some networked tests that
> appear to be a little too picky about the error message they receive
> (log below).

Neat.  I used Python 2.4.3 on Solaris 10 U3, and the tests result in the
same output.

> ERROR: test-http-proxy output changed
> --- Expected output 
> +++ Test output 
> @@ -43,4 +43,4 @@
>  added 1 changesets with 1 changes to 1 files
>  1 files updated, 0 files merged, 0 files removed, 0 files unresolved
>  %% bad host:port for proxy
> -abort: error: Connection refused
> +abort: error: Network is unreachable

I think that's a difference between errno values returned by Solaris vs.
Linux vs. BSD.  The same test `error' has been happening with Hg tests
on Solaris 10 for ages here.  I don't know what is a good/portable way
to fix this without trimming useful information out of the error.

It may be possible to filter s/\(abort: error:\) .*$/\1/ on all systems
to avoid depending on the exact errno value returned when the connection
attempt fails, but then we lose parts of the error message :(