Hi Peff, On Fri, 1 Jul 2016, Jeff King wrote: > On Fri, Jul 01, 2016 at 10:04:44AM +0200, Johannes Schindelin wrote: > > > On Fri, 1 Jul 2016, Jeff King wrote: > > > > > I waffled between the two mechanisms. Opinions welcome. > > > > I am happy you took the cmd_main() approach: we do have to play some > > tricks on Windows, in particular in some upcoming changes that support > > building with MS Visual C++ (we want to ensure that `argv` is in UTF-8, > > which means that we actually have to use the UTF-16 versions and convert > > them manually lest argv has the current Windows encoding of strings). > > Which means that we still have to use that mingw_startup() trick you > > mentioned, and which would have interfered had you chosen a similar > > method. > > I actually wondered if it would make sense to get rid of the > mingw_main() macro, and do it here as just: > > #ifdef MINGW > mingw_startup(); > #endif > > or something. But I didn't look deeply at it, and anyway I am afraid to > touch anything in that area because I can't even compile-test it. Sure. There are a couple of patches in flight to support MSVC better, and one part is a duplication of mingw_startup(). I would like to fix that before merging, of course, and hope that it will naturally be helped by your patch series. > > > diff --git a/common-main.c b/common-main.c > > > new file mode 100644 > > > index 0000000..2b96bbf > > > --- /dev/null > > > +++ b/common-main.c > > > @@ -0,0 +1,12 @@ > > > +#include "git-compat-util.h" > > > + > > > +int main(int argc, char **av) > > > +{ > > > + /* > > > + * This const trickery is explained in > > > + * 84d32bf7678259c08406571cd6ce4b7a6724dcba > > > > This commit message says that mingw_main() is declared with char **argv, > > and that is the reason why we have to convert. Maybe spell that out here, > > and then in a subsequent commit, we can fix the mingw_main() declaration? > > The description was sufficiently long that I didn't want to try > repeating it, and opted for a reference instead. If you're planning to > fix mingw_main(), I'd be happy to do that as a preparatory patch, and > then just skip this trickery entirely. :) Deal: ½/5 is in 5c54dff5c54e68a1101d8fe37aefc6158fddd7f2 and the fixup for 1/5 is in 7b74f7aabb56b428c74f5983c066dc9ea8fe5116 in the 'common-main' branch on https://github.com/dscho/git. (I had to resolve merge conflicts in a couple of the later patches, so feel free to just use the branch, but please note that I cherry-picked a patch to let me compile with DEVELOPER=1 on Windows, so you might want to drop 7b74f7aabb56b428c74f5983c066dc9ea8fe5116.) Ciao, Dscho