Potential issue with memory usage

[Adrian Buehlmann]

On 2011-05-19 09:26, Adrian Buehlmann wrote:
> What's interesting is, TortoiseHgOverlayServer is not using PyQt at all,
> just Mercurial and some weird Windows API functions. The overlay server
> doesn't even load extensions.

I'm a moron. Of course it loads extensions. mercurial/hg.py has:

def repository(ui, path='', create=False):
    """return a repository object for the specified path"""
    repo = _lookup(path).instance(ui, path, create)
    ui = getattr(repo, "ui", ui)
    for name, module in extensions.extensions():
        hook = getattr(module, 'reposetup', None)
        if hook:
            hook(ui, repo)
    return repo

So extension are in again for possible culprits.

I also just got this by private email (I hope it's ok to paste it here):

On 2011-05-19 09:37, Eduard-Cristian Stefan wrote:
> On Thu, May 19, 2011 at 10:26, Adrian Buehlmann <adrian at cadifra.com> wrote:
>> What's interesting is, TortoiseHgOverlayServer is not using PyQt at all,
>> just Mercurial and some weird Windows API functions. The overlay server
>> doesn't even load extensions.
> 
> A few months ago I had TortoiseHg (from msi installer) configured to
> load the hggit extension, which in turn used Dulwich from source. In
> order to update Dulwich files I had to kill the overlay server. Of
> course, it may not be relevant at all, since the setup used at that
> time was a little twisted.

The report at https://bitbucket.org/tortoisehg/thg/issue/702 lists the
following extensions:

<snip>
[extensions]
color =
children = 
graphlog = 
convert = 
extdiff = 
highlight = 
progress =
eol = 
keyword = 
mq = 
transplant = 
rebase = 
mercurial_keyring = 
bookmarks = 
fetch = 
#hggit=c:\Program Files (x86)\TortoiseHg\extensions\hggit
</snip>

What's telling is, that hggit is commented out.