Re: svn commit: rev 7004 - branches/svnserve-thread-pools/subversion/svnserve

Garrett Rooney wrote:

> [removed svn@subversion.tigris.org, I'm tired of getting this on both
> lists]
>
> mark benedetto king wrote:
>
>> On Mon, Sep 08, 2003 at 06:24:03PM +0200, Branko ??ibej wrote:
>>
>>> Sort of. What we need is a reliable Win32 implementation of condition
>>> variables in APR -- the current one is full of holes and races -- which
>>> is far from trivial to do. Without that, svnserve would be broken on
>>> Windows if we merged this to the trunk.
>>>
>>
>>
>> Attached is a COMPLETELY UNTESTED patch against APR's HEAD. Care to
>> take
>> a glance?
>
>
> I believe Branko posted a couple of patches to dev@apr.apache.org
> which fixed the problem, but for some reason they were never committed.
>
> Was it just lack of someone to review them?

No, the problem is those patchis *didn't* fix the problem -- they still
have race conditions. A lot less than what's currently there, but this
is an all-or-nothing situation.

mbk: After a quick glance at your patch, I can say that you're making
the same mistakes I did in my patches. :-)

> I don't recally any actual concerns being voiced...

I simply wasn't ready to commit them, because those patches are broken.
Just FYI, a correct implementation of condition variables on Windows
doesn't look even remotely like what's in APR now, or what mbk's or my
patches look like. I know a couple of places where I could get such an
implementation, but the trouble is that the licenses they're distrbuted
under makes them unusable for APR. Which means that I can't even look at
them. I have seen the algorithm described, though, so I'm (slowly)
taking it from there.

-- 
Brane Čibej   <brane_at_xbc.nu>   http://www.xbc.nu/brane/
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org