On 9/14/06, Giovanni Bajo <[EMAIL PROTECTED]> wrote: > Todd Stafford wrote: > > >> Through some research, i figured out that the popen2.popen4() call > >> errors out if the "cmd" returns more thank 4096 bytes of output. > [...] > > I was mistaken, it was 4096kb. So the actual byte count would always > > stick at 32769b. > > Thanks for the investigation. I do not know where this limit comes from. > > One thing that we could experiment with is the use of "subprocess" as an > optional alternative to popen2. subprocess.py is part of Python 2.4+, and is > backward compatible with 2.2+. Moreover, it's going to be committed to > contrib/client-side (i.e. next to svnmerge.py) in the Subversion repository, > so that makes it even more easily available. > > The idea would be to use subprocess if available (which should guarantee > also the same codepath for both Windows and Unices), and otherwise go back > to the current launch method. If you are willing to attempt and debug this > with your repository, I'd be glad to review the patch. > -- > Giovanni Bajo > > _______________________________________________ > Svnmerge mailing list > [email protected] > http://www.orcaware.com/mailman/listinfo/svnmerge
If you can send me a copy of the modified script, I can give it a try. I have Python 2.4.3 installed and it seems to have /usr/local/lib/python2.4/subprocess.py so in theory my environment should be capable of using it. I'm not sure how much debugging I can do, but I can try running against my repository which breaks with the trunk version of svnmerge.py and let you know if it fixed it or not. _______________________________________________ Svnmerge mailing list [email protected] http://www.orcaware.com/mailman/listinfo/svnmerge
