Re: pristine-tar: FTBFS on hurd-i386 (for review)
Svante Signell, le Tue 11 Feb 2014 19:49:46 +0100, a écrit :
> On Tue, 2014-02-11 at 19:24 +0100, Samuel Thibault wrote:
> > Svante Signell, le Tue 11 Feb 2014 18:52:28 +0100, a écrit :
> > > Maybe I got it wrong then, perhaps const would be better.
> >
> > Const would mean you're preventing yourself from writing to it. But you
> > *want* to write to it with snprintf. Why do you believe you need to add
> > something else than char*?
>
> Never mind for this case, (and I used static, to make it local for that
> function?)
For function-local variables, static does not change the visibility, it
only changes whether it's allocated on the stack or permanently.
> > > - char path_buf[PATH_MAX];
> > > + char *path_buf = NULL;
> >
> > Why setting it to NULL?
>
> To avoid compiler warnings?
The warning was for a good reason:
> > > - snprintf(path_buf, sizeof(path_buf), "%s/%s", PKGLIBDIR, zombie);
> > > + len = strlen(PKGLIBDIR) + 1 + strlen(zombie) + 1;
> >
> > That seems to be missing a malloc here, doesn't it?
>
> Ah, thanks!
This one. Do not consider compiler warning as noise, but as clue that
something was wrong in your code. It's very rare that the compiler is
wrong about warnings.
Samuel
Reply to: