Ping...
Has anybody had a chance to review Martin's latest patch? If nothing
happens in the next few days, I'll add it to the issue tracker.
-Hyrum
Martin Furter wrote:
> On Mon, 26 Mar 2007, Martin Furter wrote:
>> On Mon, 26 Mar 2007, David Glasser wrote:
>>> On 3/26/07, Martin Furter <mf@rola.ch> wrote:
>>>> After reading APR docs and playing around I found apr_procattr_io_set.
>>>> First I copied svn_io_start_cmd and modified it to do what I want,
>>>> and it
>>>> works fine. So I thought let's write a patch...
>>>>
>>>> Then I found out that the only caller of svn_io_start_cmd does the same
>>>> thing but in a different way. It is run_hook_cmd in
>>>> libsvn_repos/hooks.c .
>>>
>>> I don't think it makes much sense to add a new API that we're not
>>> using. Maybe you should include the change to run_hook_cmd in your
>>> patch?
>>
>> Not using yet...
>>
>> If run_hook_cmd would use the new API about 15 lines of code could be
>> removed there (creating and closing pipes).
>>
>> I stopped doing that because I wondered why it's done different there
>> duplicating code which is already in APR and works fine here (APR 1.x
>> and APR from Apache 2.0.55 on linux).
>>
>> Is there a known problem with apr_procattr_io_set ?
>>
>> If now I'll write a patch for run_hook_cmd and run the testsuite...
>
> OK, I wrote that patch, but I can't test it now because of the
> XFail(hook_test) in commit_tests.py . Will look into that later.
>
> I just thought I'll show you what would be changed.
>
> Hmm, there's another caller to apr_proc_create in
> subversion/libsvn_ra_svn/client.c: make_tunnel.
> Not sure if it makes sense to replace that with svn_io_start_cmd2 too.
>
> [[[
> Update run_hook_cmd to use svn_io_start_cmd2.
>
> * subversion/libsvn_repos/hooks.c
> (run_hook_cmd): Remove pipes and use pipe_flags of svn_io_start_cmd2.
> ]]]
>
> - Martin
Received on Tue May 8 16:30:22 2007