23.0.60; etc/DOC-${version}.buildnumber

23.0.60; etc/DOC-${version}.buildnumber

[Peter Dyballa]

Hello!

This file is not deleted by 'make clean', so upon a new build its  
number is incremented and more and more DOC files get installed. Is  
this a bug or is there a target that deletes superfluous DOC files?

--
Greetings

   Pete

Some day we may discover how to make magnets that can point in any  
direction.

Re: 23.0.60; etc/DOC-${version}.buildnumber

[Eli Zaretskii]

> From: Peter Dyballa <Peter_Dyballa@Freenet.DE>
> Date: Sat, 12 Apr 2008 14:00:23 +0200
> Cc: 
> 
> This file is not deleted by 'make clean', so upon a new build its  
> number is incremented and more and more DOC files get installed.

I think this is intended, as you still have the Emacs binary from the
previous build (emacs-${version}.buildnumber) around, and could use
it, e.g., for comparison or some other purpose.

Re: 23.0.60; etc/DOC-${version}.buildnumber

[Peter Dyballa]

Am 12.04.2008 um 15:10 schrieb Eli Zaretskii:
>> From: Peter Dyballa
>> Date: Sat, 12 Apr 2008 14:00:23 +0200
>> Cc:
>>
>> This file is not deleted by 'make clean', so upon a new build its
>> number is incremented and more and more DOC files get installed.
>
> I think this is intended, as you still have the Emacs binary from the
> previous build (emacs-${version}.buildnumber) around, and could use
> it, e.g., for comparison or some other purpose.


Not when I install or make clean. The installed Emacs wants etc/DOC-$ 
{version}.<highest buildnumber>, the cleaned (deleted) Emacs cannot  
want anything.

--
Greetings

   Pete

$ sumascii BILL GATES
   B   I   L   L   G   A   T   E   S
  66+ 73+ 76+ 76+ 71+ 65+ 84+ 69+ 83 = 663

  and add 3 because he's Bill Gates the third.

Re: 23.0.60; etc/DOC-${version}.buildnumber

[Eli Zaretskii]

> Cc: emacs-pretest-bug@gnu.org
> From: Peter Dyballa <Peter_Dyballa@Freenet.DE>
> Date: Sat, 12 Apr 2008 17:14:20 +0200
> 
> > I think this is intended, as you still have the Emacs binary from the
> > previous build (emacs-${version}.buildnumber) around, and could use
> > it, e.g., for comparison or some other purpose.
> 
> 
> Not when I install or make clean. The installed Emacs wants etc/DOC-$ 
> {version}.<highest buildnumber>, the cleaned (deleted) Emacs cannot  
> want anything.

It used to be the case that the built binary was installed under 2
different names: `emacs' and `emacs-${version}.buildnumber', which
were both hard links to the same file.  Deleting `emacs' would then
leave the other one around, and when invoked, it would want the DOC
file with the corresponding suffix.  Is this no longer the case?  That
is, does installing a new version completely erases the old one?

Re: 23.0.60; etc/DOC-${version}.buildnumber

[Peter Dyballa]

Am 12.04.2008 um 18:43 schrieb Eli Zaretskii:
> It used to be the case that the built binary was installed under 2
> different names: `emacs' and `emacs-${version}.buildnumber', which
> were both hard links to the same file.  Deleting `emacs' would then
> leave the other one around, and when invoked, it would want the DOC
> file with the corresponding suffix.  Is this no longer the case?

This is still the case.

> That is, does installing a new version completely erases the old one?


Yes.

I made an experiment. Around noon I made clean and then updated from  
CVS, finally made bootstrap. Since etc/DOC-23.0.60.1 was not removed  
(with both src/emacs and src/emacs-23.0.60.1) I complained, because  
in the weeks ago it happened that, when this DOC file survived  
deletion, a new one was created with the buildnumber increased by one  
either by make or by 'make bootstrap', which was then installed  
together with etc/emacs-${version}.buildnumber. (The weather was so  
fine and sunny that I stopped seeing a sense in deleting the file by  
hand.)

After 'make bootstrap' and then 'make clean' these files existed:

-rw-r--r--   1 root  admin  2245430  7 Apr 00:42 /usr/local/share/ 
emacs/23.0.60/etc/DOC-23.0.60.1
-rw-r--r--   1 pete  admin  2243354 12 Apr 14:58 etc/DOC-23.0.60.1

I re-compiled with a simple make and decided to install. Then these  
files existed:

-rw-r--r--   1 root  admin  2243354 12 Apr 20:07 /usr/local/share/ 
emacs/23.0.60/etc/DOC-23.0.60.1
-rw-r--r--   1 pete  admin  2243354 12 Apr 20:07 etc/DOC
-rw-r--r--   1 pete  admin  2243354 12 Apr 20:07 etc/DOC-23.0.60.1

That's really bad! This isn't the same I encountered before a few  
times! Did someone fix the bug? Or was I doing the wrong thing? So I  
decided to 'make clean' and 'make bootstrap' again. After more than  
an hour these files now exist:

-rw-r--r--   1 root  admin  2243354 12 Apr 20:07 /usr/local/share/ 
emacs/23.0.60/etc/DOC-23.0.60.1
-rw-r--r--   1 pete  admin  2243354 12 Apr 21:29 etc/DOC
-rw-r--r--   1 pete  admin  2243354 12 Apr 21:29 etc/DOC-23.0.60.1

So it seems that the case is now solved! Although etc/DOC-$ 
{version}.buildnumber is left and etc/DOC is removed, it does not  
happen any more that etc/DOC-${version}.<buildnumber + 1> is built,  
and later potentially installed. My complaint came too late ... Sorry!

--
Greetings

   Pete

For some reason, this fortune reminds everyone of Marvin Zelkowitz.

Re: 23.0.60; etc/DOC-${version}.buildnumber

[Sven Joachim]

On 2008-04-12 15:10 +0200, Eli Zaretskii wrote:

>> From: Peter Dyballa <Peter_Dyballa@Freenet.DE>
>> Date: Sat, 12 Apr 2008 14:00:23 +0200
>> Cc: 
>> 
>> This file is not deleted by 'make clean', so upon a new build its  
>> number is incremented and more and more DOC files get installed.
>
> I think this is intended, as you still have the Emacs binary from the
> previous build (emacs-${version}.buildnumber) around, and could use
> it, e.g., for comparison or some other purpose.

That's not true, the Emacs binaries are all deleted by `make clean'.
I had already reported that inconsistency some months before, see my
message in [1] and Glenn's reply in [2].

Sven

[1] http://lists.gnu.org/archive/html/emacs-devel/2007-07/msg01778.html
[2] http://lists.gnu.org/archive/html/emacs-devel/2007-08/msg00004.html

Re: 23.0.60; etc/DOC-${version}.buildnumber

[Eli Zaretskii]

> From: Sven Joachim <svenjoac@gmx.de>
> Cc: Peter Dyballa <Peter_Dyballa@Freenet.DE>,  emacs-pretest-bug@gnu.org
> Date: Sat, 12 Apr 2008 18:32:32 +0200
> 
> On 2008-04-12 15:10 +0200, Eli Zaretskii wrote:
> 
> >> From: Peter Dyballa <Peter_Dyballa@Freenet.DE>
> >> Date: Sat, 12 Apr 2008 14:00:23 +0200
> >> Cc: 
> >> 
> >> This file is not deleted by 'make clean', so upon a new build its  
> >> number is incremented and more and more DOC files get installed.
> >
> > I think this is intended, as you still have the Emacs binary from the
> > previous build (emacs-${version}.buildnumber) around, and could use
> > it, e.g., for comparison or some other purpose.
> 
> That's not true, the Emacs binaries are all deleted by `make clean'.

Maybe I completely misunderstand the issue at hand: are we talking
about files in the build directory or in the install directory?
("make clean" is relevant to the former, not the latter.)

If Peter and you are talking about the build directory, then the
previous DOC should probably be removed, but do we really want the
build number NOT to be incremented anyway?

Re: 23.0.60; etc/DOC-${version}.buildnumber

[Sven Joachim]

On 2008-04-12 19:24 +0200, Eli Zaretskii wrote:

>> From: Sven Joachim <svenjoac@gmx.de>
>> Cc: Peter Dyballa <Peter_Dyballa@Freenet.DE>,  emacs-pretest-bug@gnu.org
>> Date: Sat, 12 Apr 2008 18:32:32 +0200
>> 
>> On 2008-04-12 15:10 +0200, Eli Zaretskii wrote:
>> 
>> >> From: Peter Dyballa <Peter_Dyballa@Freenet.DE>
>> >> Date: Sat, 12 Apr 2008 14:00:23 +0200
>> >> Cc: 
>> >> 
>> >> This file is not deleted by 'make clean', so upon a new build its  
>> >> number is incremented and more and more DOC files get installed.
>> >
>> > I think this is intended, as you still have the Emacs binary from the
>> > previous build (emacs-${version}.buildnumber) around, and could use
>> > it, e.g., for comparison or some other purpose.
>> 
>> That's not true, the Emacs binaries are all deleted by `make clean'.
>
> Maybe I completely misunderstand the issue at hand: are we talking
> about files in the build directory or in the install directory?

The build directory.

> ("make clean" is relevant to the former, not the latter.)

Of course.

> If Peter and you are talking about the build directory, then the
> previous DOC should probably be removed, but do we really want the
> build number NOT to be incremented anyway?

I'm not certain I understand that; make clean removes all the emacs*
binaries from the build tree and therefore should also remove the then
useless DOC-* files.  It probably makes sense to reset the build number
to 1 after that, doesn't it?  Or what do you mean?

Sven

Re: 23.0.60; etc/DOC-${version}.buildnumber

[Eli Zaretskii]

> From: Sven Joachim <svenjoac@gmx.de>
> Cc: Peter_Dyballa@Freenet.DE,  emacs-pretest-bug@gnu.org
> Date: Sat, 12 Apr 2008 19:54:11 +0200
> 
> > If Peter and you are talking about the build directory, then the
> > previous DOC should probably be removed, but do we really want the
> > build number NOT to be incremented anyway?
> 
> I'm not certain I understand that; make clean removes all the emacs*
> binaries from the build tree and therefore should also remove the then
> useless DOC-* files.  It probably makes sense to reset the build number
> to 1 after that, doesn't it?  Or what do you mean?

I meant that perhaps the build number should grow even after "make
clean".

Re: 23.0.60; etc/DOC-${version}.buildnumber

[Peter Dyballa]

Am 12.04.2008 um 20:57 schrieb Eli Zaretskii:
> I meant that perhaps the build number should grow even after "make
> clean".

Why?

--
Greetings

   Pete

Make it simple, as simple as possible but no simpler.
				– Albert Einstein

Re: 23.0.60; etc/DOC-${version}.buildnumber

[Eli Zaretskii]

> Cc: Sven Joachim <svenjoac@gmx.de>,
>  emacs-pretest-bug@gnu.org
> From: Peter Dyballa <Peter_Dyballa@Freenet.DE>
> Date: Sat, 12 Apr 2008 21:00:37 +0200
> 
> 
> Am 12.04.2008 um 20:57 schrieb Eli Zaretskii:
> > I meant that perhaps the build number should grow even after "make
> > clean".
> 
> Why?

It's a new build, isn't it?

Re: 23.0.60; etc/DOC-${version}.buildnumber

[Peter Dyballa]

Am 12.04.2008 um 22:30 schrieb Eli Zaretskii:
> It's a new build, isn't it?


But from a new and different source code base.

--
Greetings

   Pete

These are my principles and if you don't like them... well, I have  
others.
				- Groucho Marx

Re: 23.0.60; etc/DOC-${version}.buildnumber

[Eli Zaretskii]

> Cc: svenjoac@gmx.de,
>  emacs-pretest-bug@gnu.org
> From: Peter Dyballa <Peter_Dyballa@Freenet.DE>
> Date: Sat, 12 Apr 2008 22:59:26 +0200
> 
> 
> Am 12.04.2008 um 22:30 schrieb Eli Zaretskii:
> > It's a new build, isn't it?
> 
> 
> But from a new and different source code base.

No, "make clean" by itself doesn't change the sources.

Re: 23.0.60; etc/DOC-${version}.buildnumber

[Peter Dyballa]

Am 13.04.2008 um 05:18 schrieb Eli Zaretskii:
>> ut from a new and different source code base.
>
> No, "make clean" by itself doesn't change the sources.


Right, it's me who prepares with a 'make clean' the CVS update to  
enter a new state. I want to have clean basis before I enter a new  
cycle. And I don't care what number this cycle has (reaching 100  
after one year?).

Anyway, 'make clean' expresses something like: "make way for  
something new." This new does not necessarily depend on new input,  
like my apartment is still the same when I have swept away all the  
debris that fell down from my body and clothes etc. This 'make clean'  
makes place for something new. If I want to have some counter  
incremented I could 'make increment.'

When I make clean I don't want my rent be incremented.

--
Greetings

   Pete
                       ~  o
                        ~_\\_/\
                       ~  O   O

Re: 23.0.60; etc/DOC-${version}.buildnumber

[Eli Zaretskii]

> Cc: svenjoac@gmx.de,
>  emacs-pretest-bug@gnu.org
> From: Peter Dyballa <Peter_Dyballa@Freenet.DE>
> Date: Sun, 13 Apr 2008 10:38:08 +0200
> 
> Anyway, 'make clean' expresses something like: "make way for  
> something new."

That's not my interpretation of "make clean".  It just means "return
to the state we had before configure+make".  Also, I don't think "make
clean" was designed for anything close to the situation where the
files are updated en masse from a VCS.  Sounds like you want something
like "make cvs-clean" or some such.

Re: 23.0.60; etc/DOC-${version}.buildnumber

[Stephen J. Turnbull]

Eli Zaretskii writes:

 > I meant that perhaps the build number should grow even after "make
 > clean".

Yes, it should, unless "make clean" reverts all the source changes
that might affect DOC.  That "cleaned" build might very well have been
installed to preserve it for later comparisons.  Then it is very
useful to have a convenient unique identifier for each build.

Re: 23.0.60; etc/DOC-${version}.buildnumber

[Peter Dyballa]

Am 12.04.2008 um 22:49 schrieb Stephen J. Turnbull:
> Then it is very useful to have a convenient unique identifier for  
> each build.

GNU Emacs has this information when you start to send a bug report.


Upon configuration some files are overwritten. Upon 'make bootstrap'  
some more files are overwritten even twice. Upon 'make install' all  
files (hopefully) are overwritten. Isn't that already enough uniqueness?

--
Greetings

   Pete

To most people solutions mean finding the answers. But to chemists  
solutions
are things that are still all mixed up.

Re: 23.0.60; etc/DOC-${version}.buildnumber

[Stephen J. Turnbull]

Peter Dyballa writes:
 > 
 > Am 12.04.2008 um 22:49 schrieb Stephen J. Turnbull:
 > > Then it is very useful to have a convenient unique identifier for  
 > > each build.
 > 
 > GNU Emacs has this information when you start to send a bug report.
 > 
 > 
 > Upon configuration some files are overwritten. Upon 'make bootstrap'  
 > some more files are overwritten even twice. Upon 'make install' all  
 > files (hopefully) are overwritten. Isn't that already enough uniqueness?

No.  Because the files are overwritten, you cannot determine whether
the builds are the same or not.

Re: 23.0.60; etc/DOC-${version}.buildnumber

[Stefan Monnier]

> I'm not certain I understand that; make clean removes all the emacs*
> binaries from the build tree and therefore should also remove the then
> useless DOC-* files.  It probably makes sense to reset the build number
> to 1 after that, doesn't it?  Or what do you mean?

Yes, that makes sense.

Note that nowadays it's not even clear there's much point in having
DOC-<vers>: we could just use a single DOC file: Emacs is able to
gracefully react to a DOC file that has been changed.

I've been using such a setup for several years now, and  occasionally it
fails to notice that the file has changed and ends up showing some other
docstring than the one intended, but this is rather rare (it happened to
me recently and I felt like "Wow, so the practice does agree with the
theory after all").


        Stefan

Re: 23.0.60; etc/DOC-${version}.buildnumber

[Peter Dyballa]

Am 12.04.2008 um 19:24 schrieb Eli Zaretskii:
> Maybe I completely misunderstand the issue at hand: are we talking
> about files in the build directory or in the install directory?
> ("make clean" is relevant to the former, not the latter.)

About both. Because the DOC files in the build tree are also installed.

>
> If Peter and you are talking about the build directory, then the
> previous DOC should probably be removed, but do we really want the
> build number NOT to be incremented anyway?

I want to have control over this mechanism. Usually I want to have  
only one build number version (and I don't want to waste time by re- 
configuring again after a 'make distclean' or such). Because when it  
happens that I start over, none of the installed superfluous DOC  
files and Emacs versions are de-installed. This worked until some day  
in winter, in February, I think. Then the etc/DOC-$ 
{version}.buildnumber<s> files were saved and new ones with  
incremented build numbers were created, and sometimes also installed  
when I did not pay attention. This seems to have changed recently (I  
am still waiting for a 'make bootstrap' to finish) ...

--
Greetings

   Pete

Behold the warranty ... the bold print giveth and the fine print  
taketh away.