Hi Bruno,

Bruno Haible <[email protected]> writes:

> Thanks Collin, for having reported the issue to the musl list.

No problem.

> The test failure is still present in today's CI, on Alpine Linux.

Yes, I haven't made any changes because of it. The only realistic
solution I see is adding the following snippet to the start of each
program:

    #if MUSL_LIBC
      if (isatty (STDOUT_FILENO))
        setvbuf (stdout, NULL, _IOFBF, BUFSIZ);
    #endif

Given the number of programs, I obviously do not love that solution.

After defining what "fully buffered", etc. mean POSIX says [1]:

    Support for these characteristics is implementation-defined, and may
    be affected via setbuf() and setvbuf().

So I guess Musl is compliant, technically. But I find the behavior a bit
silly.

Collin

[1] 
https://pubs.opengroup.org/onlinepubs/9799919799/functions/V2_chap02.html#tag_16_05

  • misc/io-errors fails on ... Bruno Haible via GNU coreutils General Discussion
    • Re: misc/io-errors ... Collin Funk
      • Re: misc/io-err... Bruno Haible via GNU coreutils General Discussion
        • Re: misc/io... Collin Funk
          • Re: mis... Bruno Haible via GNU coreutils General Discussion
            • Re... Collin Funk
              • ... Bruno Haible via GNU coreutils General Discussion
                • ... Bruno Haible via GNU coreutils General Discussion
                • ... Collin Funk
                • ... Bruno Haible via GNU coreutils General Discussion
                • ... Collin Funk
                • ... Bruno Haible via GNU coreutils General Discussion
                • ... Collin Funk
                • ... Bruno Haible via GNU coreutils General Discussion

Reply via email to