rawhide report: 20050315 changes

[Iago Rubio]

On Tue, 2005-03-15 at 11:03 -0600, Brian Millett wrote:
> On Tue, 2005-03-15 at 17:55 +0100, Iago Rubio wrote:
> > On Tue, 2005-03-15 at 09:55 -0600, Brian Millett wrote:
> > > Had to make some cosmetic patches to get it to compile with gcc4 and I
> > > am sure they are "wrong" but it compiled.
> > > 
> > > [bpm]$ diff -wruN nm-netlink-monitor.c~ nm-netlink-monitor.c
> > > --- nm-netlink-monitor.c~       2005-03-14 23:30:14.000000000 -0600
> > > +++ nm-netlink-monitor.c        2005-03-15 09:39:23.000000000 -0600
> > > @@ -568,10 +568,10 @@
> > >                                   NmNetlinkMonitor *monitor)
> > >  {
> > >         GError *error;
> > > -       gchar *received_bytes;
> > > +       gchar *received_bytes="";
> > >         gboolean processing_is_done;
> > > -       gsize num_received_bytes;
> > > -       guint num_bytes_to_process;
> > > +       gsize num_received_bytes=0;
> > > +       guint num_bytes_to_process=0;
> > >         struct nlmsghdr *header;
> > 
> > Frankly, I don't see how this could fix the build.
> > 
> > May be it's being used -Werror in the gcc command line ?
> 
> Why yes it is. (from Makefile)
> CC = gcc
> CCDEPMODE = depmode=gcc3
> CFLAGS = -Wall -Werror -std=gnu89 -g -O2 -Wno-unused
> -Wno-strict-aliasing -Wno-sign-compare -Wdeclaration-after-statement
> -Wno-pointer-sign
> CPP = gcc -E
> 
> So I do not know what or why the difference.

-Werror will stop compilation on warnings.

I think it should be better to take out the -Werror and let the sources
untouched, as to randomly initialize variables could make things fail.

Taking out -Werror, you will be able to compile, and will let the
warnings inform other developers about this problem.

Initializing the variables this way you achieve the opposite of what the
developer wanted. He wanted to care of small glitches so he decided to
convert warnings in error.

You're blindly hiding the warnings, and it can make debugging more
painful.

In the case you don't care about this, I recommend at least initialize
also the gchar* variable to 0.

gchar *received_bytes=0;

This way if it's g_freed anywhere - at least - it won't segfault.

Just my 2 - euro - cents.
-- 
Iago Rubio