Point hgsubversion to self-compiled subvertpy? (Windows)

[Steve Borho]

On 08/22, Judson Craft wrote:
> On Friday, August 22, 2014 at 12:11 PM, Steve Borho wrote:
> > This doesn't necessarily answer your question, but it's probably an
> > alternative solution.
> > 
> > FWIW: I've never found functioning subvertpy libraries for Windows 
> 
> I assumed it was in more common use... The mercurial documentation I
> read itself seemed to recommend subvertpy over swig-python (even
> though of course those are the official SVN bindings), so I put the
> effort into doing it manually in lieu of having a good binary package
> in PyPI. I've compiled subvertpy for Python 32-bit bindings
> successfully on Windows 7 with Visual Studio 2008.
> 
> Now I need to specifically recompile everything for a 64-bit Python
> environment on Windows, since one repository I'm working with is
> exhausting memory at ~1GB working memory. I know 32-bit apps can grow
> up to 3-4GB on Linux before hitting OOM but apparently on Win32 memory
> address space is much more limited.

Which operation is running out of memory? 

If it is a clone, you can usually work around this by using repeated
pulls with chunks of the revision history (the SVN bindings have memory
leaks) Once it is in Mercurial, the x86/x64 question is less critical,
and you can use 64bit Mercurial without the subversion bindings.

> At any rate I finally built the entire stack of svn, bindings, and
> subvertpy from scratch for 64-bit on Windows, but for some reason I
> can't get hgsubversion to detect it.
> 
> Not sure how I'm botching the installation, since the last step is to
> just run setup.py and let distutils take care of it.
> 
> I know it sounds like amateur hour here, any pointers much
> appreciated.

Starting a 64bit python shell and trying to import the subvertpy
bindings might be helpful.

-- 
Steve Borho