Compiling Weston with musl-libc (WAS: [PATCH weston] weston-launch: use custom error function)

[Murray Calavera]

On 30 September 2016 at 12:59, Eric Engestrom <eric.engestrom at imgtec.com>
wrote:

> On Fri, Sep 30, 2016 at 11:36:24AM +0100, Murray Calavera wrote:
> > On 30 September 2016 at 11:10, Eric Engestrom <eric.engestrom at imgtec.com
> >
> > wrote:
> >
> > > On Thu, Sep 29, 2016 at 09:26:16PM +0100, Murray Calavera wrote:
> > > > error.h is a gnu extension and not available in other
> > > > popular libcs like musl. This patch provides a replacement.
> > > >
> > > > Signed-off-by: Murray Calavera <murray.calavera at gmail.com>
> > >
> > > How did you test this? For me, `CC=musl-gcc ./autogen.sh` stops on:
> > >   [...]
> > >   checking for library containing pam_open_session... no
> > >   configure: error: weston-launch requires pam
> > >
> >
> > Have you got libpam installed?
> > I don't see how this patch could have affected the configure,
> > does it configure without this patch?
>
> I do have it, and it works fine with both `CC=gcc` and `CC=clang`, but
> `CC=musl-gcc` and `CC=musl-clang` both fail with that error.
>
> Is this also how you use musl?  If it is, then the issue is probably on
> my side, I won't take anymore of your time with this :)
>

​I have a native musl toolchain. The musl-gcc wrapper modifies
include and library paths​ so perhaps that is the issue.
Also what Armin said.

>
> >
> > >
> > > The code looks good though (with one nit-pick), so even if I couldn't
> > > test it, it is:
> > > Reviewed-by: Eric Engestrom <eric.engestrom at imgtec.com>
> > >
> > > > ---
> > > >  libweston/weston-launch.c | 20 +++++++++++++++++++-
> > > >  1 file changed, 19 insertions(+), 1 deletion(-)
> > > >
> > > > diff --git a/libweston/weston-launch.c b/libweston/weston-launch.c
> > > > index 140fde1..84f7d60 100644
> > > > --- a/libweston/weston-launch.c
> > > > +++ b/libweston/weston-launch.c
> > > > @@ -33,7 +33,6 @@
> > > >  #include <poll.h>
> > > >  #include <errno.h>
> > > >
> > > > -#include <error.h>
> > > >  #include <getopt.h>
> > > >
> > > >  #include <sys/types.h>
> > > > @@ -112,6 +111,25 @@ struct weston_launch {
> > > >
> > > >  union cmsg_data { unsigned char b[4]; int fd; };
> > > >
> > > > +static void
> > > > +error(int status, int errnum, const char *msg, ...)
> > > > +{
> > > > +     va_list args;
> > > > +
> > > > +     fputs("weston-launch: ", stderr);
> > > > +     va_start(args, msg);
> > > > +     vfprintf(stderr, msg, args);
> > > > +     va_end(args);
> > > > +
> > > > +     if (errnum)
> > > > +             fprintf(stderr, ": %s\n", strerror(errnum));
> > > > +     else
> > > > +             fputc('\n', stderr);
> > >
> > > Why not `fprintf(stderr, "\n");`?
> > > While fputc() is enough since this is a single char, the use of
> > > a different function here looks... odd.
> > >
> >
> > As you said, because I'm not printing formatted data
> > there is no need to use printf.
> > However if the consensus here is to use printf even
> > when not needed, I can change that.
>
> I just thought it looked odd, but I have no real argument one way or the
> other, so feel free to leave it as is :)
>
> Cheers,
>   Eric
>
> >
> > >
> > > > +
> > > > +     if (status)
> > > > +             exit(status);
> > > > +}
> > > > +
> > > >  static gid_t *
> > > >  read_groups(void)
> > > >  {
> > > > --
> > > > 2.10.0
> > >
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/wayland-devel/attachments/20160930/4c77e7a7/attachment.html>