Hi Greg,
Greg Hudson wrote:
>On Sat, 2004-05-29 at 13:04, Jostein Chr. Andersen wrote:
>
>
>>If we want to have a "real" service, then we must edit the existing code
>>for svnserve.
>>I will probably manage to do this, but it will take long time (my c
>>knowledge is limited, as an example, I used 3 days when writing my first
>>usable C program: packages/win32-innosetup/tools/svnpath/main.c, and
>>that program do still have some issues).
>>
>>
>
>This isn't going to help, but: I'm going to pretty much insist that code
>of this nature go into APR behind a nice clean interface, not into
>svnserve itself. The amount of non-Subversion logic in svnserve should
>be minimized. Just as not every APR-using Unix daemon should have to
>implement all the Unix-specific daemon logic (although it's only a few
>lines of code), not every APR-using Windows service should have to
>implement all the Windows-specific service logic, especially since it
>seems to bottom out at a hundred lines of code or so. (And Apache seems
>to spend almost 1500 lines of code on it.)
>
>
>
The 1400 lines of apache code deal with running on win9x/me and on
winnt/2k/xp/2k3. Since running svnserve on win9x/me is not an option
anyway, that would mean only half or less of the code for svnserve.
Putting that code into apr would either mean one has to separated the
module into 3 parts (win9x, winnt and switching code) or we have to link
to the win9x code, which we do not need.
Could any of our apr committers say if moving the code from the
mpm/winnt to apr is an option.
Patrick
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Sun May 30 07:10:04 2004