Re: Require GCC 4.3 or later to build glibc

["Carlos O'Donell" <carlos at systemhalted dot org>]

On Fri, Feb 24, 2012 at 12:43 PM, Joseph S. Myers
<joseph@codesourcery.com> wrote:
> On Fri, 24 Feb 2012, Roland McGrath wrote:
>
>> > It appears current practice is not to mention regeneration of generated
>> > files (such as INSTALL and configure) in ChangeLogs. ?We could discuss
>> > changing that to something more like GCC and Binutils, where the logs do
>> > say "Regenerate." for generated files.
>>
>> That was always the convention ("Regenerated.") and nobody ever discussed
>> changing it. ?Some people have failed to observe it in the last few months.
>
> I had been creating such entries (through habit from GCC and elsewhere),
> but:
>
> * The last time INSTALL is mentioned in the ChangeLogs is 2006-03-06 when
> you did the 2.4 release. ?And there have been plenty of install.texi
> changes since then. ?Likewise plenty of configure.in changes without
> configure regeneration noted.
>
> * When I checked in a change with such an entry, it was shortly thereafter
> followed by:
>
> commit 6c1eace5df647ea85c7472c3fb314665abb22ec1
> Author: Ulrich Drepper <drepper@gmail.com>
> Date: ? Mon Jan 30 22:50:20 2012 -0500
>
> ? ?Don't add CL for generated files!!!
>
> which I took as indicating an explicit glibc convention against such
> entries that I had failed to notice (and while I think it's better to have
> such ChangeLog entries, conventions for such things as Bugzilla usage and
> release management were certainly more important things to agree on).

I think we should have ChangeLog entries for regenerated files that
say "Regenerated."

It shows the author consciously knew that file had to be regenerated.

For example I will admit that I didn't have a clue that INSTALL was
auto-generated.

In fact this whole discussion struck off guard owing to my own stupidity :-)

Cheers,
Carlos.