This is the mail archive of the
mailing list for the glibc project.
Re: Manual varies by machine.
- From: "Carlos O'Donell" <carlos at redhat dot com>
- To: "Joseph S. Myers" <joseph at codesourcery dot com>
- Cc: GNU C Library <libc-alpha at sourceware dot org>
- Date: Mon, 15 Sep 2014 20:48:40 -0400
- Subject: Re: Manual varies by machine.
- Authentication-results: sourceware.org; auth=none
- References: <5415B4F1 dot 4000307 at redhat dot com> <Pine dot LNX dot 4 dot 64 dot 1409151555240 dot 16736 at digraph dot polyomino dot org dot uk>
On 09/15/2014 11:59 AM, Joseph S. Myers wrote:
> On Sun, 14 Sep 2014, Carlos O'Donell wrote:
>> The present glibc manual varies by machine. Would it be
>> acceptable to make it a general principle that the manual
>> should *not* vary by machine and that it should always be
>> consistent for any build?
> It should not vary by machine. (Of course it may vary depending on
> Texinfo version. If there is nondeterminism in makeinfo output, that
> should be fixed there in line with general principles of reproducible
I found the problem, and it was indeed a defect in my version of
libm-err-tab.pl which caused it to be different.
I agree they should be the same, I believed it might not be
deterministic, but it is.
>> The most glaring variations are because of libm-err-tab.pl
>> being generated per machine. I'd hope we could check something
> libm-err-tab.pl should not cause the manual to vary by machine, because it
> should always generate output tables that cover all machines.
Agreed. I'll make sure that's always the case.
>> in for the manual there. Not to mention I worry about the
>> license of the resulting output of that file.
> I don't think that's a meaningful question; I doubt these factual tables
> of errors in libm functions have sufficient expressive content to be
> eligible for copyright.