On Fri, Oct 25, 2013 at 08:26:29PM -0400, Jeff King wrote:

> On Fri, Oct 25, 2013 at 04:06:19PM -0700, Junio C Hamano wrote:
> > Jeff King <p...@peff.net> writes:
> > 
> > > diff --git a/pack-bitmap.c b/pack-bitmap.c
> > > new file mode 100644
> > > index 0000000..73c52fd
> > > --- /dev/null
> > > +++ b/pack-bitmap.c
> > > @@ -0,0 +1,965 @@
> > > +#include <stdlib.h>
> > > +
> > > +#include "cache.h"
> > 
> > You among all people already know why this is bad, no?
> Yes, I am well aware of why we do not want it. I thought I removed that,
> but it appears I missed one. Sorry.

In addition to that one, there are a few other system header includes:

  - the ewah/*.c files include the necessary standard headers, and do
    not include git-compat-util.h at all. I do not really consider these
    "git code", but rather a black box we have imported into the tree
    to ease the dependency chain. So in that sense, they operate like
    xdiff/*.c or compat/regex/*.c, which also compile on
    their own (and can get away with it because they are mostly standard
    C and do not do "system" things.

    However, the code in the ewah/ directory has been hacked up a bit
    from its original, and ewah_io.c _does_ include "git-compat-util.h".
    So it may make sense to consider our copy a fork and git-ify it
    more. The upstream Java EWAH library is active and separate, but the
    C port done by Vicent, while theoretically a separate project, does
    not actually have any users besides git (though presumably libgit2
    will build on it at some point).

  - Regardless of the decision above, ewah/ewok.h includes stdlib.h,
    which will pollute git source files which use it.

  - Ditto for khash.h.

To unsubscribe from this list: send the line "unsubscribe git" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Reply via email to