Re: [kbd] [PATCH 5/5] Replace error() by fprintf() and perror()

[Felix Janda <felix.janda@posteo.de>]

Alexey Gladkov wrote:
> 22.04.2015 22:06, Felix Janda пишет:
> > ---
> >  src/kbdinfo.c        | 26 ++++++++++++++++----------
> >  src/setvtrgb.c       | 42 ++++++++++++++++++++++++------------------
> >  src/vlock/auth.c     |  2 --
> >  src/vlock/pam.c      |  6 ++----
> >  src/vlock/username.c | 12 ++++++++----
> >  src/vlock/vlock.c    | 11 ++++++-----
> >  src/vlock/vt.c       |  5 ++---
> >  7 files changed, 58 insertions(+), 46 deletions(-)
> 
> Applied, except for the last one. Please check out the master branch.

Thanks for applying the other patches and for kbd_error()! Now everything
compiles fine with musl libc.

However kbd_error() does not behave correctly in the case that the first
argument is 0 (EXIT_SUCCESS). error() does _not_ terminate the program
if the first argument is EXIT_SUCCESS. Maybe have in addition a
kbd_warn() so that the __attribute__((noreturn)) can stay for kbd_error()
(then the exitnum argument would also become superfluous).

My excuse [1] for changing the error messages (no more program name) was
because the source is inconsistent. For example init_vt() in vlock/vt.c
or main() in kbdinfo.c mix fprintf(stderr) and error(). It seems to me
that error was mainly used as a convenient shorthand for
fprint(stderr) && exit(). (Having a shorthand makes sense.)

Thanks,
Felix

PS: Could you maybe note on the webpage that git.altlinux.org is the
main git repository and might be ahead of the git.kernel.org repo?

[1]: http://lists.altlinux.org/pipermail/kbd/2014-December/000487.html