a fix for the numerous errors "differ in signedness" with GCC4 ?

[stranche@…]

I found this interesting article ​http://forum.openvz.org/index.php?t=msg&goto=1265&

It may be usefull to correct the numerous errors "differ in signedness" when
compiling with GCC4.
Here I paste the content (the page may diseapper as well).

Hi,

today after building my openvz development environment i just wanted to
build the userspace tools like vzquota and vzctl. I tried to build
vzquota with the default Makefile and CFLAGS which includes '-Werror'
GCC4 complains about the following:

cc1: warnings being treated as errors
quotaon.c: In function 'quotaugidset_proc':
quotaon.c:1228: warning: pointer targets in passing argument 2 of 'str2uint'
differ in signedness
make[1]: * [quotaon.o] Error 1

Attached is a small fix which only changes the 'int ugid' (which is default
signed) to 'unsigned int ugid' Nothing more. :)

--Maik

diff -Nur vzquota-2.7.0-7/src/quotaon.c vzquota-2.7.0-7-signedness-fix/src/quotaon.c
--- vzquota-2.7.0-7/src/quotaon.c 2005-09-12 17:46:07.000000000 +0200
+++ vzquota-2.7.0-7-signedness-fix/src/quotaon.c 2006-02-04 16:37:30.233562250 +0100
@@ -1221,7 +1221,7 @@
quota_setugid(0, NULL, &info);
} else {
struct dq_stat stat;

• int ugid;

+ unsigned int ugid;

/* get ugid */
if (argc != 5

[diego@…]

Thanks for the hint, but we are aware of how to fix those warnings. If you wish
to see them gone, send a patch ...

[stranche@…]

Okay people,

Here is a tree cleaned of (hopefully) every "warning: pointer targets in passing
argument * of '*' differ in signedness" when compiling on GCC4.

Tree took from CVS on 21 feb.

rather than a huge diff that may not be the best way to see what changes have
been done, I made 2 tgz files.
One with the original tree I got from CVS, one with the tree corrected.
So that you still can do diffs if you like.

get it at :
​http://stranche.free.fr/mplayer-orig.tgz
​http://stranche.free.fr/mplayer-cleaned.tgz

[diego@…]

Dominik, over to you, these are compiler warnings ...

[Dominik 'Rathann' Mierzejewski]

Please provide the "huge diff".

[stranche@…]

diff between the original tree

[Dominik 'Rathann' Mierzejewski]

Did I mention it should be a unified diff?
​http://www.mplayerhq.hu/DOCS/tech/patches.txt
Besides, your patch contains unrelated changes, like:

diff -r main/config.mak main2/config.mak
76c76
< SRC_PATH=/home/stranche/test/main
---

SRC_PATH=/home/stranche/sauv-mplayer/main

diff -r main/DOCS/man/en/CVS/Entries main2/DOCS/man/en/CVS/Entries
1c1
< /mplayer.1/1.1231/Mon Feb 20 09:36:10 2006
---

/mplayer.1/1.1232/Mon Feb 20 23:50:40 2006

diff -r main/DOCS/man/en/mplayer.1 main2/DOCS/man/en/mplayer.1
1c1
< .\" $Revision: 1.1231 $
---

.\" $Revision: 1.1232 $

And so on. Please clean it up before sending it again.

[stranche@…]

diff done with 'diff -Naur' and tested with 'patch -p0'

[reimar]

Except maybe for a few select cases, this just makes the code ugly.
Disabling the warning would be the _much_ better solution IMHO.

[Dominik 'Rathann' Mierzejewski]

I agree with Reimar. And since some of these touch libavcodec, I suggest posting
them to ffmpeg-devel. I think most of these casts won't be accepted, but those
hunks which change signed<->unsigned variables just might.

[stranche@…]

(In reply to comment #9)

Are you in contact directly with the ffmpeg people or should I propose the patch
myself ?
Anyway, I did all this to fix at least a part of the numerous warnings that made
the compilation ugly on GCC4, maybe it was not the most "clean" way to do but it
can give a few hints to make it right then.
After all, in GCC3 it is not reported at all but in GCC4 it is recognized, maybe
in the future such warning could become fatal.

[stranche@…]

(In reply to comment #8)

You mean using something like -Wno-sign-compare ?

[diego@…]

Much of this no longer applies and/or is bogus.