[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: Subversion & Windows

From: Ben Reser <ben_at_reser.org>
Date: Tue, 8 Jan 2013 13:52:32 -0800

On Tue, Jan 8, 2013 at 1:30 PM, Johan Corveleyn <jcorvel_at_gmail.com> wrote:
> I think the problem is worse with 1.6. At least I myself was
> absolutely not looking forward to testing another release of 1.6 (but
> did it anyway, after a week or so of waiting). As you say, setting up
> a build environment for Subversion on Windows is already pretty hard.
> But then, if you have your build env finetuned for building trunk
> and/or 1.7, going back to build 1.6 is a real pain:

Right that was a typo I meant 1.6 was worse.

> - Can't build out of the box with VS 2010, because gen-win.py on 1.6.x
> only generates up to 2008 format (so you either have to do it with VS
> 2008 if you still have it (I didn't ... recently uninstalled it), or
> you "convert" it from 2008 format to 2010 format from within the VS
> 2010 GUI (which takes a long time). I know, maybe I could have
> backported the necessary stuff to 1.6.x, but ... I didn't.
>
> - Can't build with latest serf on Windows, as I noted in the signing
> thread. That took me a couple of iterations before reverting back to
> serf 0.7.1.

If we do another 1.6 release (I'm suspecting we won't) maybe we should
backport this stuff? However, it seems counterproductive to not
backport things that make it exceedingly painful for us to do a
release. Going forward maybe we should consider these sorts of things
for backport even on our supported but not recommended releases.

> - Running the test suite with 1.6 takes a looooong time on Windows.
> It's at least twice as slow as the 1.7 testsuite on my laptop. I start
> a .bat file calling the 4 combinations in sequence (local, svn, neon,
> serf) before going to sleep, and in the morning it's still only
> halfway the neon run (with the serf run still coming up). This is
> probably mostly a testament to the performance improvement that WC-NG
> brought to 1.7, especially for Windows/NTFS. But then it's doubly
> painful to see this slow 1.6 running on my laptop ... ouch that hurts,
> don't want to go there again :-).

Good to know that has gotten better with newer releases. I noticed
how slow it was when I did it.

> I think maybe we could lower the required number of sigs for 1.6 (or
> for whichever release is the one-before-eol'ed version). It should be
> very stable by now, the backports are very conservative and there are
> few of them, and probably the number of users that will install it is
> also declining...

Maybe this is something we should only do for 1.6.x due to the
difficulties. Given that we may not release another 1.6.x we may not
even need to use the lesser vote procedure.
Received on 2013-01-08 22:53:10 CET

This is an archived mail posted to the Subversion Dev mailing list.

This site is subject to the Apache Privacy Policy and the Apache Public Forum Archive Policy.