threads / patch / 15209

patchFix a warning (on cygwin) to allow -Werror

Subject: [PATCH] Fix a warning (on cygwin) to allow -Werror

## tl;dr

4 messages between Aug 26, 2008 and Aug 29, 2008. Diffs are folded; open one to read it.

replies: 3people: 2as markdown or json

Ramsay Jones· Aug 26, 2008, 17:50 UTC · lore

Signed-off-by: Ramsay Jones <ramsay@ramsay1.demon.co.uk>
---

Noticed while building 1.6.0 from the release tar-ball.

Also, I noticed that git_ntohl() is still defined in compat/mingw.h; is this still required? It's not a problem, just curious! (I don't have mingw installed, so I can't test myself)

ATB,

Ramsay Jones

 builtin-fast-export.c |    3 ++-
 1 files changed, 2 insertions(+), 1 deletions(-)

Show changes to builtin-fast-export.c +2 −1

diff --git a/builtin-fast-export.c b/builtin-fast-export.c
index 0709716..7c93eb8 100644
--- a/builtin-fast-export.c
+++ b/builtin-fast-export.c
@@ -417,7 +417,8 @@ static void export_marks(char *file)
 	for (i = 0; i < idnums.size; i++) {
 		if (deco->base && deco->base->type == 1) {
 			mark = ptr_to_mark(deco->decoration);
-			fprintf(f, ":%u %s\n", mark, sha1_to_hex(deco->base->sha1));
+			fprintf(f, ":%"PRIu32" %s\n", mark,
+				sha1_to_hex(deco->base->sha1));
 		}
 		deco++;
 	}

-- 
1.6.0

Johannes Sixt· Aug 27, 2008, 07:57 UTC · re: Ramsay Jones · lore

Re: [PATCH] Fix a warning (on cygwin) to allow -Werror

Ramsay Jones schrieb:

> Also, I noticed that git_ntohl() is still defined in compat/mingw.h; is this
> still required?  It's not a problem, just curious! (I don't have mingw
> installed, so I can't test myself)

Yes, it's still needed. The reason is that on Windows ntohl() returns 'unsigned long', and then gcc warns about

sha1_file.c:700: warning: unsigned int format, u_long arg (arg 3)

git_ntohl() is the workaround.

-- Hannes

Ramsay Jones· Aug 28, 2008, 16:09 UTC · re: Johannes Sixt · lore

Re: [PATCH] Fix a warning (on cygwin) to allow -Werror

Johannes Sixt wrote:

Show 11 quoted lines

> Ramsay Jones schrieb:
>> Also, I noticed that git_ntohl() is still defined in compat/mingw.h; is this
>> still required?  It's not a problem, just curious! (I don't have mingw
>> installed, so I can't test myself)
> 
> Yes, it's still needed. The reason is that on Windows ntohl() returns
> 'unsigned long', and then gcc warns about
> 
> sha1_file.c:700: warning: unsigned int format, u_long arg (arg 3)
> 
> git_ntohl() is the workaround.

OK, Thanks.

Ah, yes, everything works on cygwin and Linux because:

                             Linux           Cygwin
    ntohl() returns:         uint32_t        uint32_t
    uint32_t typedef-ed as:  unsigned int    unsigned long
    PRIu32 #define-d as:     "u"             "lu"

So, I suspect (IOW I am guessing that) mingw must have:

    ntohl() returns:         unsigned long (well u_long anyway)
    uint32_t typedef-ed as:  unsigned int
    PRIu32 #define-d ad:     "u"

(I've just noticed that you are not #include-ing inttypes.h, so you are relying on the PRIu32 defined in git-compat-util.h)

If the above is true, then well, ... Bummer ;-)

I suppose a (very) slight improvement would be to change the return type of your git_ntohl() wrapper to uint32_t. (But we are talking very slight ;-)

ATB,

Ramsay Jones

Johannes Sixt· Aug 29, 2008, 06:23 UTC · re: Ramsay Jones · lore

Re: [PATCH] Fix a warning (on cygwin) to allow -Werror

Ramsay Jones schrieb:

> (I've just noticed that you are not #include-ing inttypes.h, so you are
> relying on the PRIu32 defined in git-compat-util.h)

We have to: MinGW include files are not C99.

-- Hannes