On Thursday 31 March 2011, Ralf Wildenhues wrote: > * Stefano Lattarini wrote on Thu, Mar 31, 2011 at 03:27:18PM CEST: > > On Wednesday 30 March 2011, Ralf Wildenhues wrote: > > > > > > A general comment: if you keep on adding tests to maint, it will break > > > branch-1.11 in the long run, in the sense that the branch will not be > > > releasable any more without splitting up the test suite, or improving > > > the test suite driver as discussed before, as it will not work any more > > > on MinGW/MSYS. Please keep this in mind. > > > > > How much room is left, approximately? > > I haven't tested on MinGW recently, and haven't tested branch-1.11 in a > longer while. The limit was 32K for argv plus environment IIRC. > But then we should be pretty safe on both maint and master: $ (cd tests && echo *.test | wc -c) 11154 # maint 16089 # master Even considering the extra code in the Makefile rules, these values are waaay below the upper length limit of MSYS/MinGW. Then, why have you experienced failures due to command-line length on MSYS/MinGW? I'm confused ...
> > I have no big problems in applying > > this patch to master only, but I'd rather keep it in maint unless we are > > already dangerously near to the MinGW/MSYS upper limit. > > Well, apply it based on maint, > The patch is already based on maint ... but then, should I apply it to maint, or merge it to master only? > and let's strive for leaving rougly 6K for tests we may still want to > introduce, plus environment. Regards, Stefano