Excerpts from Yuya Nishihara's message of 2016-08-12 09:26:37 +0900: > On Thu, 11 Aug 2016 18:28:09 +0100, Jun Wu wrote: > > Excerpts from Yuya Nishihara's message of 2016-08-12 01:55:44 +0900: > > > Maybe we'll need setproctitle() equivalent for Linux which will update > > > argv? > > > > That would be probably rewriting argv[0], afaik. I think it is impossible > > to get the "argv" pointer in a reasonable way. > > Er, Py_GetArgcArgv() ? No idea if we can reliably use this hidden API. > > https://hg.python.org/cpython/file/v2.7.11/Modules/main.c#l691
Nice find! I did see "static char **orig_argv;" but missed the comment around it. I'll try the approach. The memory handling could be a bit tricky. One reason I was using prctl is because it is well defined on Linux, with the annoying 15-char limit. argv hack is not portable and may have compatibility issues with other strange code. I will double check with MRIRuby. > > > 16-char limitation of prctl() would come from pthread API. If we want to > > > extend the (ab)use of process title to hgweb, PR_SET_NAME won't be useful. > > > It won't be displayed in normal ps output. > > > > It's still useful with '-O comm', and atop will record it. prctl is lower > > level than pthread. Sadly 16 is hardcoded in Linux. > > I assume '-O comm' would be less common for Linux sysadmins. _______________________________________________ Mercurial-devel mailing list Mercurial-devel@mercurial-scm.org https://www.mercurial-scm.org/mailman/listinfo/mercurial-devel