Changelog time format

Changelog time format

[Martin von Loewis]

For Emacs users, it would be convenient if the following patch is
applied to the various ChangeLog files, most prominently
gcc/ChangeLog.

Thanks,
Martin

Index: gcc/ChangeLog
===================================================================
RCS file: /egcs/carton/cvsfiles/egcs/gcc/ChangeLog,v
retrieving revision 1.1261
diff -u -p -r1.1261 ChangeLog
--- gcc/ChangeLog	1998/05/08 21:48:51	1.1261
+++ gcc/ChangeLog	1998/05/09 13:40:12
@@ -8929,3 +8954,7 @@ Mon Aug 11 10:04:49 1997  Jeffrey A Law 
 Sun Aug 10 12:00:20 1997  Jeffrey A Law  (law@cygnus.com)
 
 	* egcs project officially starts.
+
+Local Variables:
+add-log-time-format: current-time-string
+End:

Re: Changelog time format

[Jeffrey A Law]

  In message < 199805091349.PAA00458@mira.isdn.cs.tu-berlin.de >you write:
  > For Emacs users, it would be convenient if the following patch is
  > applied to the various ChangeLog files, most prominently
  > gcc/ChangeLog.
What does this patch actually do?

If it's supposed to make ChangeLog formats all the same, it's
probably worth doing, even though some of us use vi macros to
build ChangeLog entries :-)

jeff

Re: Changelog time format

[Martin von Loewis]

> What does this patch actually do?

It sets a variable on the ChangeLog buffer. The default of this
variable is `add-log-iso8601-time-string', at least for Emacs 20.
With this default setting, M-x add-change-log-entry will give me
a header of

1998-05-10  Martin von Loewis  <martin@mira.isdn.cs.tu-berlin.de>

	* 

I believe this default was chosen because ISO 8601 says it looks
better :-) With this setting, I get

Sun May 10 09:59:42 1998  Martin von Loewis  <martin@mira.isdn.cs.tu-berlin.de>

	* 

There is still some non-uniformity in the time formats, as some people
put the time zone in, and others don't (the Emacs ChangeLog mode
thinks that the time zone is part of the developer's name). However,
this difference is hardly noticable.

Martin

Re: Changelog time format

[Jim Wilson]

I noticed that this was installed, but it doesn't seem to do anything useful.

Simply adding it to one of the ChangeLog files in gcc isn't interesting.
We either add it to all of them or none of them.  We can't add it to all
of them unless we get everyone to agree that the old format is better.
I doubt that we can get everyone to agree on the old format.  The FSF
would not have started using a new format by default unless they wanted
everyone to use it.

Also, the change isn't going to force all ChangeLog entries to use the
new format.  I always create a patch file, add the ChangeLog entry to the
patch file, and then copy it into the ChangeLog file.  Hence all of the
patches I create on my Linux machine will continue to use the new format.

Jim

Re: Changelog time format

[Jeffrey A Law]

  In message < 199805172212.PAA04043@rtl.cygnus.com >you write:
  > I noticed that this was installed, but it doesn't seem to do anything usefu l.
It's of marginal value (IMHO).   But I didn't see a signifcant
reason not to install the change.

If folks object, we can always pull it out.


  > Simply adding it to one of the ChangeLog files in gcc isn't interesting.
  > We either add it to all of them or none of them.
Well, 99% of our changes go into gcc/ChangeLog and gcc/cp/ChangeLog

And (IMHO) we probably don't want to go adding this to things like
texinfo, libio, libstdc++ which are maintained externally.



  >  We can't add it to all
  > of them unless we get everyone to agree that the old format is better.
I don't think we need to get everyone to agree.  We can't please
everyone, so we shouldn't try

  > Also, the change isn't going to force all ChangeLog entries to use the
  > new format.  I always create a patch file, add the ChangeLog entry to the
  > patch file, and then copy it into the ChangeLog file.  Hence all of the
  > patches I create on my Linux machine will continue to use the new format.
Absolutely.  I actually make my ChangeLog entries with a vi macro,

jeff

Re: Changelog time format

[Andreas Schwab]

Jim Wilson <wilson@cygnus.com> writes:

|> I noticed that this was installed, but it doesn't seem to do anything useful.
|> Simply adding it to one of the ChangeLog files in gcc isn't interesting.
|> We either add it to all of them or none of them.  We can't add it to all
|> of them unless we get everyone to agree that the old format is better.
|> I doubt that we can get everyone to agree on the old format.  The FSF
|> would not have started using a new format by default unless they wanted
|> everyone to use it.

|> Also, the change isn't going to force all ChangeLog entries to use the
|> new format.  I always create a patch file, add the ChangeLog entry to the
|> patch file, and then copy it into the ChangeLog file.  Hence all of the
|> patches I create on my Linux machine will continue to use the new format.

IMHO the best solution would be to install a wrapper in CVS that converts
the time stamps to the old format on checkin.  The only problem is that
the new format lacks timeofday information.

-- 
Andreas Schwab                                      "And now for something
schwab@issan.informatik.uni-dortmund.de              completely different"
schwab@gnu.org

Re: Changelog time format

[Franz Sirl]

At 02:21 18.05.98 , Jeffrey A Law wrote:
>
>  In message < 199805172212.PAA04043@rtl.cygnus.com >you write:
>  > I noticed that this was installed, but it doesn't seem to do anything
usefu l.
>It's of marginal value (IMHO).   But I didn't see a signifcant
>reason not to install the change.
>
>If folks object, we can always pull it out.

I would vote for a common format, it's easier to read. I would prefer the
standard email notation for the timezone like +0200 or -0800.

What do you think about a Make-rule which modifies version.c with the
ChangeLog CVS revision numbers (gcc/cp/f) as soon as the CVS/Entries dir
structure is detected? I think this will fulfill the need for exact bug
reporting at little cost (especially as soon as Alexandre's reporting
scripts are checked in) and you don't have to maintain anything, CVS and
make do it for you.

Bye,
Franz.

Re: Changelog time format

[Jim Wilson]

	IMHO the best solution would be to install a wrapper in CVS that converts
	the time stamps to the old format on checkin.  The only problem is that
	the new format lacks timeofday information.

I like to think of that as a feature.  I find it annoying that the current
formats have timeofday info, but do not include timezone info, and hence
they can be misleading.  If you really need timeofday info, get it from
the CVS logs.

The timeofday info is also inaccurate because there is often little
correlation between the time of the changelog entry and the time that
the patch is actually checked in.

I think the best solution is to simply accept the fact that there are two
formats for ChangeLog entries, and that the new format will prevail over
time because the FSF has adopted it as the standard (i.e. default) format.

If you want to argue the issue, then argue it with the FSF.  EGCS should
just be following the FSF style guidelines, not creating our own.

Jim

Re: Changelog time format

[Martin von Loewis]

> If you want to argue the issue, then argue it with the FSF.  EGCS should
> just be following the FSF style guidelines, not creating our own.

IMHO, it does not matter that much what the format is, as long as it
is uniform. I find the current mix of two (and a half) formats
distracts me when reading the changes.

There is a coding style enforced on source code. It would be in GNU
tradition to enforce one on ChangeLog entries, too.

Martin

Re: Changelog time format

[Ulrich Drepper]

wilson@cygnus.com (Jim Wilson) writes:

> If you want to argue the issue, then argue it with the FSF.  EGCS should
> just be following the FSF style guidelines, not creating our own.

Many GNU packages (tar, recode, gettext, glibc, sharutils, patch,
fileutils, grep) are already using the ISO 8601 format and for many
others it's simply a matter of getting out a new version.  gcc/egcs
are again one of the last packages to convert to this format.

-- Uli
---------------.      drepper at gnu.org  ,-.   1325 Chesapeake Terrace
Ulrich Drepper  \    ,-------------------'   \  Sunnyvale, CA 94089 USA
Cygnus Solutions `--' drepper at cygnus.com   `------------------------

Re: Changelog time format

[Jim Wilson]

	Many GNU packages (tar, recode, gettext, glibc, sharutils, patch,
	fileutils, grep) are already using the ISO 8601 format and for many
	others it's simply a matter of getting out a new version.  gcc/egcs
	are again one of the last packages to convert to this format.

I think you are ignoring the fact that gcc/egcs are fundamentally different
than the other packages you mentioned.  Compare how many people have write
access to glibc versus how many people have write access to egcs.  The more
people are involved, the longer it takes to make process changes.  It only
took one person to decide that glibc should switch to the new format.  It
will take 10 or so for egcs to decide to switch to the new format.  That
is why egcs is inevitably lagging behind.

It would actually help egcs if the ChangeLog format was explicitly written
into the FSF style guides.  Then at least my arguments in favor of the
new format would be more convincing.

Jim

Re: Changelog time format

[Ulrich Drepper]

Jim Wilson <wilson@cygnus.com> writes:

> It only took one person to decide that glibc should switch to the
> new format.  It will take 10 or so for egcs to decide to switch to
> the new format.  That is why egcs is inevitably lagging behind.

It took only one person to decide, right, but they are fortunately
several contributors who send patches with ChangeLog entries which
should be written in the right form (and some of the contributors have
write access).  It's not so much different, it's just that nobody
wants to stand up and decide.

> It would actually help egcs if the ChangeLog format was explicitly written
> into the FSF style guides.  Then at least my arguments in favor of the
> new format would be more convincing.

If you want to discuss this with Richard, go on.  I've given up doing
things like this and we (some other GNU maintainers) have written our
own extensions to the existing GNU coding standard which are more
detailed and more strict.

-- Uli
---------------.      drepper at gnu.org  ,-.   1325 Chesapeake Terrace
Ulrich Drepper  \    ,-------------------'   \  Sunnyvale, CA 94089 USA
Cygnus Solutions `--' drepper at cygnus.com   `------------------------

Re: Changelog time format

[Dave Love]

In case people using older Emacsen are interested, the Emacs 20
add-log.el (which implements the new ChangeLog format) should be
usable easily in a recent Emacs 19.  It just needs a stub for the
`custom' stuff, easily supplied.

Re: Changelog time format

[Ulrich Drepper]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 1063 bytes --]

Jim Wilson <wilson@cygnus.com> writes:

> I also think it is a waste of my time to reformat/reject ChangeLog
> entries from people who haven't installed emacs-20 yet, and hence
> can't generate the proper kind of ChangeLog entry

This can be automated using a simple Emacs macro.  François Pinard
also has a Perl script which changes entire ChangeLog in one run (I
should have to lying around).

> I hadn't realized that alternative GNU coding standards were available.
> Where would I find this document?

I don't think we have it written down all the way.  The major result
of all this was the automake tool which generates Makefiles as they
should be.  Some of us tried to start writing all this down but didn't
finished it due to lack of time.  But since we review each others
packages there is a coding standard which is followed.

-- Uli
---------------.      drepper at gnu.org  ,-.   1325 Chesapeake Terrace
Ulrich Drepper  \    ,-------------------'   \  Sunnyvale, CA 94089 USA
Cygnus Solutions `--' drepper at cygnus.com   `------------------------

Re: Changelog time format

[Jim Wilson]

	It's not so much different, it's just that nobody
	wants to stand up and decide.

I am certainly willing to stand up and make a decision.  However, I have not
done so because I don't think it is appropriate.  I still believe that the
egcs situation is rather different than say the glibc situation.

I also think it is a waste of my time to reformat/reject ChangeLog entries from
people who haven't installed emacs-20 yet, and hence can't generate the
proper kind of ChangeLog entry

If think if we just wait patiently, egcs will eventually go where we want it
to go.

	I've given up doing
	things like this and we (some other GNU maintainers) have written our
	own extensions to the existing GNU coding standard which are more
	detailed and more strict.

I hadn't realized that alternative GNU coding standards were available.
Where would I find this document?

Jim

Re: Changelog time format

[Mike Stump]

> To: Andreas Schwab <schwab@issan.informatik.uni-dortmund.de>
> Date: Mon, 18 May 1998 12:44:42 -0700
> From: Jim Wilson <wilson@cygnus.com>

> If you want to argue the issue, then argue it with the FSF.  EGCS should
> just be following the FSF style guidelines, not creating our own.

Then we should put in the below patch.


Doing diffs in gcc/cp/ChangeLog.~1~:
*** gcc/cp/ChangeLog.~1~	Tue May 19 10:56:26 1998
--- gcc/cp/ChangeLog	Tue May 19 11:03:21 1998
*************** Wed Oct 11 16:30:34 1995  Brendan Kehoe 
*** 9578,9587 ****
  
  	* parse.y (fn.def1): Call split_specs_attrs in
  	declmods notype_declarator case.
- \f
- Use a consistent time stamp format in ChangeLog entries.
- Not everyone has Emacs 20 yet, so stick with Emacs 19 format for now.
- 
- Local Variables:
- add-log-time-format: current-time-string
- End:
--- 9578,9580 ----
--------------