GCC build failed for native with your patch on 2003-05-12T16:15:13Z.

[Geoffrey Keating]

On Monday, May 12, 2003, at 01:29  PM, Benjamin Kosnik wrote:

>
>> This was a real failing build, it was caused by Jan's patch to
>> output_constant in varasm.c.
>> and the data is in the form year-month-day.
>
> There is no way to tell this from the quoted ChangeLog.
>
> Hasn't this been broken since the first ChangeLog date? Ie, for two 
> months now?

Yes, we've had various compiler regressions since 2003-03-03T21:26:39Z, 
which IMO is pretty terrible and I wish people would put more effort 
into testing their patches and fixing their regressions.  However, 
right at the top of the produced ChangeLog diff, you'll notice it says 
something like:

> ChangeLog entries since last run on 2003-05-12T18:32:50Z:
> --- /Users/regress/tbox/changelog_mail/gcc/gcc/ChangeLog	Mon May 12 
> 11:33:58 2003
> +++ gcc/gcc/ChangeLog	Mon May 12 11:57:48 2003
> @@ -1,3 +1,9 @@
> +2003-05-12  Mark Mitchell  <mark@codesourcery.com>
> +
> +	PR other/10745
> +	* configure.in: Correct detection of GNU ld version number.
> +	* configure: Regenerated.
> +
>  2003-05-12  Zack Weinberg  <zack@codesourcery.com>
>
>  	* diagnostic.c (diagnostic_for_decl): Take a
>
> ChangeLog entries between 2003-05-12T18:32:50Z and the last time 
> everything
> passed, on 2003-03-03T21:26:39Z:
> [lots o' stuff]

That is, the first few entries are the changes since the last run (your 
ChangeLog entry would have appeared in there), and the rest are older 
entries since the last time there were no regressions.

> If not, you might want to trim the quoted ChangeLog to better reflect
> reality about what is causing the breakage. It's being truncated 
> already
> because it has too much info.

Well, somewhere in there are all the ChangeLog entries that reflect the 
breakage.  It's hard to do better in an automated system.

-- 
Geoff Keating <geoffk@apple.com>