Re: Implementing notrans_man_MANS

2008-03-12 Thread Peter Breitenlohner 

On Tue, 11 Mar 2008, Ralf Wildenhues wrote:


Having one test is fine, and it looks good, too.

I suppose we might want to change the MANS line and the install-manX
rules to be flattened, they already look pretty ugly in the Makefile.
But that can happen in a followup patch.


Hello Ralf,

I guess that could mean a rule starting with
install-man6: dist_man_MANS man_MANS nodist_man_MANS
notrans_dist_man_MANS notrans_man_MANS \
notrans_nodist_man_MANS notrans_nodist_man_MANS

If so, that would make
grep '^install-man6:' Makefile.in | grep '\$(notrans_nodist_man_MANS)'
in the present notrans.test FAIL.

Maybe it would then be better that the notrans.test script creates none of
the files and have them all created by make rules. (At one time I was not
sure if Automake complains if some of these files are missing.)

To really test these dependencies one could even do:
rm -f foo.1 z-baz.man
$MAKE install-man1
rm -f bar.2 z-bar.man
$MAKE install-man2
and so on.

Regards,
Peter

Re: Implementing notrans_man_MANS

2008-03-10 Thread Peter Breitenlohner 

On Thu, 6 Mar 2008, Ralf Wildenhues wrote:


* Peter Breitenlohner wrote on Thu, Mar 06, 2008 at 01:27:05PM CET:

or still something else. If so please indicate what.


Oh sorry, I was in a hurry and not thinking.  What I meant was that
while you can't wrap arbitrary lines ending in backslash-newline in `if
%?NOTRANS_MANS%/else/endif', you can do so at a finer grainage than
whole rules: it's just that it won't work after lines ending in
backslash.

To some extent this is independent of the question you ask about the
rule target names: whether you interleave trans and notrans in the .am
file and whether you give trans and notrans different target names need
not be dealt with at the same time.  So please for now leave you patch
as is, but only change all those NOTR/DOTR prefixes to use if/else/endif
(which includes a few lines of duplication in the .am file).

Clearer now?  If not, then I guess I can also patch that later.


Hi Ralf,

yes, thanks. However, I can only use if/endif if/endif and not if/else/endif
because a section may have trans_ as well as notrans_ man pages.

=

Attached are revised versions of the three patches. I hope they take care of
all your earlier comments.

patch-03-notrans_MANS-impl:

The new variables names are more or less self explanatory (I hope).

I had to use
  @unsorted_deps = (keys %notrans_vars, keys %trans_vars,
keys %notrans_this_sect, keys %trans_this_sect);
  my @deps = sort @unsorted_deps;
instead of
  my @deps = sort (keys %notrans_vars, keys %trans_vars,
   keys %notrans_this_sect, keys %trans_this_sect);
because this may give a perl warning [sort (...) used as function],
resulting in three test (instsh.tes, location.test, warnopts.test) to FAIL.

patch-04-notrans_MANS-doc:

I have slightly reworded the text preceding the example, and hope this
clarifies the intent.

patch-05-notrans_MANS-test:

notrans.test now contains all combinations. I have incorporated the checks
that used to be in man4.test into notrans.test, because otherwise there
would have been much repetition.  Hope this is OK.  If so, remember not to
mention man4.test in the ChangLog.  If not, I could make that two separate
tests.

The install-man depedndencies are checked either directly (inspecting
Makefile.in) or indirectly (letting make generate the files).

=

regards
Peter Breitenlohner [EMAIL PROTECTED]diff -ur -N -x 'automake.info*' -x version.texi -x stamp-vti 
automake-1.10.1.orig/automake.in automake-1.10.1/automake.in
--- automake-1.10.1.orig/automake.in2008-02-19 18:07:41.0 +0100
+++ automake-1.10.1/automake.in 2008-03-08 00:10:34.0 +0100
@@ -3363,27 +3363,39 @@
   # Find all the sections in use.  We do this by first looking for
   # standard sections, and then looking for any additional
   # sections used in man_MANS.
-  my (%sections, %vlist);
+  my (%sections, %notrans_sections, %trans_sections,
+  %notrans_vars, %trans_vars, %notrans_sect_vars, %trans_sect_vars);
   # We handle nodist_ for uniformity.  man pages aren't distributed
   # by default so it isn't actually very important.
+  foreach my $npfx ('', 'notrans_')
+{
   foreach my $pfx ('', 'dist_', 'nodist_')
{
  # Add more sections as needed.
  foreach my $section ('0'..'9', 'n', 'l')
{
- my $varname = $pfx . 'man' . $section . '_MANS';
+ my $varname = $npfx . $pfx . 'man' . $section . '_MANS';
  if (var ($varname))
{
  $sections{$section} = 1;
  $varname = '$(' . $varname . ')';
- $vlist{$varname} = 1;
+ if ($npfx eq 'notrans_')
+   {
+ $notrans_sections{$section} = 1;
+ $notrans_sect_vars{$varname} = 1;
+   }
+ else
+   {
+ $trans_sections{$section} = 1;
+ $trans_sect_vars{$varname} = 1;
+   }
 
  push_dist_common ($varname)
if $pfx eq 'dist_';
}
}
 
- my $varname = $pfx . 'man_MANS';
+ my $varname = $npfx . $pfx . 'man_MANS';
  my $var = var ($varname);
  if ($var)
{
@@ -3393,28 +3405,87 @@
  if (/\.([0-9a-z])([a-z]*)$/)
{
  $sections{$1} = 1;
+ if ($npfx eq 'notrans_')
+   {
+ $notrans_sections{$1} = 1;
+   }
+ else
+   {
+ $trans_sections{$1} = 1;
+   }
}
}
 
  $varname = '$(' . $varname . ')';
- $vlist{$varname} = 1;
+ if ($npfx eq 'notrans_')
+   {
+

Re: Implementing notrans_man_MANS

2008-03-06 Thread Peter Breitenlohner 

On Wed, 5 Mar 2008, Ralf Wildenhues wrote:


Well, here's a review of your patches.  If you have time to address the
comments, that would be great, otherwise I will eventually do it.


Hi Ralf,

I will go through your comments and think most of them pose no problem. For
the moment just a few remarks/questions:


This triple loop is already at 72 iterations.  Not good for performance.
If this grows further, we may have to access variables in a different
manner here.


I assume you are worried by the total number of 72 iterations, not by the
fact that they come from three nested loops.  Given the Automake
infrastructure (to the extent I understand it) I see no alternative.


+?NOTRY?l2='%NOTRZ%'; \
+?NOTRY?for i in $$l2; do \
...
+?NOTR?   $(INSTALL_DATA) $$file $(DESTDIR)$(man%SECTION%dir)/$$inst; \
+?NOTR? done


Note to self: should rewrite the above when applying the multi-file
install.


What is multi-file install? Maybe I don't need to know.


Also, the whole thing would probably be a lot more readable if the
?NOTR?/?DOTR? prefixes vanished and were replaced by full rules, wrapped
in `if %?NOTRANS_MANS%'.  That would be a better name for NOTR, and
since it would then only appear a couple of times, it could be longer
without pain, too.


I have tried to wrap the make rule fragments by something like
`if %?XXX%' but failed, maybe I got the syntax wrong.  However, from
some comments elsewhere I concluded that at present this is not possible.

Do you mean something like:
install-man: install-man1 install-notrans-man1
install-man1: trans_DEPENDENCIES
trans_RULES
install-notrans-man1: notrans_DEPENDENCIES
notrans_RULES

That should not be too difficult. Shall I try it?


[EMAIL PROTECTED] must be specified first when used in conjunction with
+either @samp{dist_} or @samp{nodist_} (@pxref{Dist}).  For instance:


Can this limitation be lifted (without making the code much slower)?


I wouldn't know how, and there is an analogous limitation elsewhere:
nobase_dist_pkgdata_DATA


+./configure --program-prefix=gnu- --prefix `pwd`/inst --mandir `pwd`/inst/man


Please quote instances of `pwd' for the master testsuite.


Like --prefix `pwd`/inst for Windows paths containing spaces? I just
copied this from transform.test. A quick grep shows there are still plenty
of unquoted pwd's around.

BTW: Which one is better: `--prefix PREFIX' or `--prefix=PREFIX'?

==

In the meantime I have some ideas how to handle manpage translations. Mainly
an Autoconf/Automake macro and Makefile.am stuff, plus a little bit of help
from Automake.  To whom should I send this, when finished to write it up?

Regards Peter

Re: Implementing notrans_man_MANS

2008-03-06 Thread Ralf Wildenhues 
Hi Peter,

* Peter Breitenlohner wrote on Thu, Mar 06, 2008 at 10:43:20AM CET:
 On Wed, 5 Mar 2008, Ralf Wildenhues wrote:

 This triple loop is already at 72 iterations.  Not good for performance.
 If this grows further, we may have to access variables in a different
 manner here.

 I assume you are worried by the total number of 72 iterations, not by the
 fact that they come from three nested loops.  Given the Automake
 infrastructure (to the extent I understand it) I see no alternative.

Your assumption is correct, and yes, I don't see a simple alternative
either.

 +?NOTRY?l2='%NOTRZ%'; \
 +?NOTRY?for i in $$l2; do \
 ...
 +?NOTR?   $(INSTALL_DATA) $$file 
 $(DESTDIR)$(man%SECTION%dir)/$$inst; \
 +?NOTR? done

 Note to self: should rewrite the above when applying the multi-file
 install.

 What is multi-file install? Maybe I don't need to know.

You don't need to know, but I meant this:
http://thread.gmane.org/gmane.comp.sysutils.automake.patches/2945/focus=3088

 Also, the whole thing would probably be a lot more readable if the
 ?NOTR?/?DOTR? prefixes vanished and were replaced by full rules, wrapped
 in `if %?NOTRANS_MANS%'.  That would be a better name for NOTR, and
 since it would then only appear a couple of times, it could be longer
 without pain, too.

 I have tried to wrap the make rule fragments by something like
 `if %?XXX%' but failed, maybe I got the syntax wrong.  However, from
 some comments elsewhere I concluded that at present this is not possible.

It is possible but only if you wrap whole rules.  You cannot wrap parts.
See depend2.am for examples.

 Do you mean something like:
   install-man: install-man1 install-notrans-man1
   install-man1: trans_DEPENDENCIES
   trans_RULES
   install-notrans-man1: notrans_DEPENDENCIES
   notrans_RULES

No.

 [EMAIL PROTECTED] must be specified first when used in conjunction with
 +either @samp{dist_} or @samp{nodist_} (@pxref{Dist}).  For instance:

 Can this limitation be lifted (without making the code much slower)?

 I wouldn't know how, and there is an analogous limitation elsewhere:
 nobase_dist_pkgdata_DATA

Ah ok, I wasn't aware of that.

 +./configure --program-prefix=gnu- --prefix `pwd`/inst --mandir 
 `pwd`/inst/man

 Please quote instances of `pwd' for the master testsuite.

 Like --prefix `pwd`/inst for Windows paths containing spaces? I just
 copied this from transform.test. A quick grep shows there are still plenty
 of unquoted pwd's around.

Not if you look in git master of Automake.

 BTW: Which one is better: `--prefix PREFIX' or `--prefix=PREFIX'?

Both are accepted.

 In the meantime I have some ideas how to handle manpage translations. Mainly
 an Autoconf/Automake macro and Makefile.am stuff, plus a little bit of help
 from Automake.  To whom should I send this, when finished to write it up?

Respective bits to auto{conf,[EMAIL PROTECTED], please.

Thanks,
Ralf

Re: Implementing notrans_man_MANS

2008-03-06 Thread Ralf Wildenhues 
* Peter Breitenlohner wrote on Thu, Mar 06, 2008 at 01:27:05PM CET:
 On Thu, 6 Mar 2008, Ralf Wildenhues wrote:

 Also, the whole thing would probably be a lot more readable if the
 ?NOTR?/?DOTR? prefixes vanished and were replaced by full rules, wrapped
 in `if %?NOTRANS_MANS%'.  That would be a better name for NOTR, and
 since it would then only appear a couple of times, it could be longer
 without pain, too.

 Do you mean something like:
 install-man: install-man1 install-notrans-man1
 install-man1: trans_DEPENDENCIES
 trans_RULES
 install-notrans-man1: notrans_DEPENDENCIES
 notrans_RULES

 No.

 what then? Maybe this:

 if only trans
   install-man1: trans_DEPENDENCIES
   trans_RULES
 as before, if only notrans
   install-man1: notrans_DEPENDENCIES
   notrans_RULES
 or if both are present
   install-man1: trans_DEPENDENCIES notrans_DEPENDENCIES
   trans_RULES
   notrans_RULES

 Pro: much simpler to read
 Con: duplication of code, harder to maintain

 or still something else. If so please indicate what.

Oh sorry, I was in a hurry and not thinking.  What I meant was that
while you can't wrap arbitrary lines ending in backslash-newline in `if
%?NOTRANS_MANS%/else/endif', you can do so at a finer grainage than
whole rules: it's just that it won't work after lines ending in
backslash.

To some extent this is independent of the question you ask about the
rule target names: whether you interleave trans and notrans in the .am
file and whether you give trans and notrans different target names need
not be dealt with at the same time.  So please for now leave you patch
as is, but only change all those NOTR/DOTR prefixes to use if/else/endif
(which includes a few lines of duplication in the .am file).

Clearer now?  If not, then I guess I can also patch that later.

Thanks,
Ralf

Re: Implementing notrans_man_MANS

2008-03-05 Thread Ralf Wildenhues 
Hello Peter,

* Peter Breitenlohner wrote on Wed, Feb 20, 2008 at 02:20:04PM CET:
 
 attached are five patches (the diffs are actually against the result of
 applying your patch for PR 516 from 2008-01-22):

Thanks for your work in this, and sorry for the delay.  I saw that your
paperwork is now complete, thanks!

 2. patch-03-notrans_MANS-impl:
 
 The actual implementation of `notrans_'. The patch looks awful but the
 logic in the resulting patched file should be reasonably clear.

Indeed.  ;-)

 At the same time the dependencies for `install-man%SECTION%' have been
 revised. All make variables not used in Makefile.am are omitted from the
 install/uninstall rules. All those used in the rules also appear as
 dependencies.
 
 One consequence: make dist_man_MANS='bar.1 baz.5' install-man will only
 work well if `dist_man_MANS' is used in Makefile.am and Automake has
 generated the relevant (!notrans_) part of the install-man1 and
 install-man5 rules. So ... ?

Yeah, that kind of thing is to be expected from automake-active
variables.  Not new in this patch.

 Please feel free to revise the wording in texinfo, comments, and/or
 ChangeLog.

Well, here's a review of your patches.  If you have time to address the
comments, that would be great, otherwise I will eventually do it.

 diff -ur -N -x 'automake.info*' -x version.texi -x stamp-vti 
 automake-1.10.1.orig/automake.in automake-1.10.1/automake.in
 --- automake-1.10.1.orig/automake.in  2008-02-19 18:07:41.0 +0100
 +++ automake-1.10.1/automake.in   2008-02-20 01:52:26.0 +0100
 @@ -3363,27 +3363,38 @@
# Find all the sections in use.  We do this by first looking for
# standard sections, and then looking for any additional
# sections used in man_MANS.
 -  my (%sections, %vlist);
 +  my (%sections, %nsections, %tsections, %nvlist, %tvlist, %nsvlist, 
 %tsvlist);

Variable naming seems, uhm, a bit hard to read, starting from ugly names
and putting more on top.  ;-)

# We handle nodist_ for uniformity.  man pages aren't distributed
# by default so it isn't actually very important.
 +  foreach my $npfx ('', 'notrans_')
 +{
foreach my $pfx ('', 'dist_', 'nodist_')
   {
 # Add more sections as needed.
 foreach my $section ('0'..'9', 'n', 'l')

This triple loop is already at 72 iterations.  Not good for performance.
If this grows further, we may have to access variables in a different
manner here.

   {
 -   my $varname = $pfx . 'man' . $section . '_MANS';
 +   my $varname = $npfx . $pfx . 'man' . $section . '_MANS';
[...]

 --- automake-1.10.1.orig/lib/am/mans.am   2008-02-20 09:38:10.0 
 +0100
 +++ automake-1.10.1/lib/am/mans.am2008-02-20 00:50:53.0 +0100
 @@ -16,6 +16,20 @@
  ## Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
  ## 02110-1301, USA.
  
 +## Parameters:
 +##   SECTION string  man section
 +##   DEPSstring  dependencies
 +##
 +##   NOTRboolhave notrans_ MANS
 +##   NOTRX   string  notrans_ man%SECTION%_MANS (if any)
 +##   NOTRY   boolhave notrans_ man_MANS
 +##   NOTRZ   string  notrans_ man_MANS
 +##
 +##   DOTRboolsame
 +##   DOTRX   stringas above
 +##   DOTRY   boolwithout
 +##   DOTRZ   stringnotrans_

What is the rationale for NOTRY, do the X, Y, Z suffixes have any
meaning?  I understand NO and DO prefixes.  So what about using
   NOTR_SECT_MANS instead of  NOTRX
   NOTR_NOSECT_MANS   instead of  NOTRZ
consequently
   NOTR_NSinstead of  NOTRY

and likewise DOTR*?  For NOTR and DOTR see below.

  man%SECTION%dir = $(mandir)/man%SECTION%
  
  ##  ##
 @@ -29,37 +43,65 @@
  ?INSTALL-MAN?install-data-am: install-man
  ?INSTALL-MAN?am__installdirs += $(DESTDIR)$(man%SECTION%dir)
  .PHONY install-man: install-man%SECTION%
 -install-man%SECTION%: $(man%SECTION%_MANS) $(man_MANS)
 +install-man%SECTION%: %DEPS%
   @$(NORMAL_INSTALL)
   test -z $(man%SECTION%dir) || $(MKDIR_P) 
 $(DESTDIR)$(man%SECTION%dir)
 - @list='$(man%SECTION%_MANS) $(dist_man%SECTION%_MANS) 
 $(nodist_man%SECTION%_MANS)'; \
 +## Handle MANS with notrans_ prefix
 +?NOTR?   @list='%NOTRX%'; \
 +## Extract all items from notrans_man_MANS that should go in this section.
 +## This must be done dynamically to support conditionals.
 +?NOTRY?  l2='%NOTRZ%'; \
 +?NOTRY?  for i in $$l2; do \
 +?NOTRY?case $$i in \
 +## Have to accept files like `foo.1c'.
 +?NOTRY?  *.%SECTION%*) list=$$list $$i ;; \
 +?NOTRY?esac; \
 +?NOTRY?  done; \
 +?NOTR?   for i in $$list; do \
 +## Find the file.
 +?NOTR? if test -f $$i; then file=$$i; \
 +?NOTR? else file=$(srcdir)/$$i; fi; \
 +## Change the extension if needed.
 +?NOTR? ext=`echo $$i | sed -e 's/^.*\\.//'`; \
 +?NOTR? case $$ext in \
 +?NOTR?   %SECTION%*) ;; \
 +?NOTR?   *) ext='%SECTION%' ;; \
 +?NOTR? esac; \
 +## Extract basename of

Re: Implementing notrans_man_MANS

2008-02-20 Thread Peter Breitenlohner 

On Tue, 19 Feb 2008, Ralf Wildenhues wrote:


How should I proceed, once all that is finished? Email to you and/or
automake bugzilla?


automake-patches@gnu.org is best.


Are unified diffs against the autoamke-1.10.1 tarball OK?


Yes.


What about re-indenting existing code?


Please as a separate patch if possible, and please following the
already-used style.


Hallo Ralf,

attached are five patches (the diffs are actually against the result of
applying your patch for PR 516 from 2008-01-22):

1. patch-02-notrans_MANS-reindent:

Reindentation of a section in automake.in, anticipating an additional
foreach loop (it was less work to do this first).

2. patch-03-notrans_MANS-impl:

The actual implementation of `notrans_'. The patch looks awful but the
logic in the resulting patched file should be reasonably clear.

At the same time the dependencies for `install-man%SECTION%' have been
revised. All make variables not used in Makefile.am are omitted from the
install/uninstall rules. All those used in the rules also appear as
dependencies.

One consequence: make dist_man_MANS='bar.1 baz.5' install-man will only
work well if `dist_man_MANS' is used in Makefile.am and Automake has
generated the relevant (!notrans_) part of the install-man1 and
install-man5 rules. So ... ?

3. patch-04-notrans_MANS-doc:

Texinfo documentation for the notrans_ prefix.

4. patch-05-notrans_MANS-test:

Two new test cases for missing dependencies and the notrans_ prefix.

5. automake-1.10.1-ChangeLog:

Proposed ChangeLog entry.

==

Please feel free to revise the wording in texinfo, comments, and/or
ChangeLog.

regards
Peter Breitenlohner [EMAIL PROTECTED]Prepare for implementation of notrans_man_MANS: anticipate
an additional foreach loop and reindent accordingly.

Note: `diff -b' yields nothing.

diff -ur -N automake-1.10.1.orig/automake.in automake-1.10.1/automake.in
--- automake-1.10.1.orig/automake.in2008-01-21 23:11:41.0 +0100
+++ automake-1.10.1/automake.in 2008-02-19 18:07:41.0 +0100
@@ -3366,42 +3366,42 @@
   my (%sections, %vlist);
   # We handle nodist_ for uniformity.  man pages aren't distributed
   # by default so it isn't actually very important.
-  foreach my $pfx ('', 'dist_', 'nodist_')
-{
-  # Add more sections as needed.
-  foreach my $section ('0'..'9', 'n', 'l')
+  foreach my $pfx ('', 'dist_', 'nodist_')
{
- my $varname = $pfx . 'man' . $section . '_MANS';
- if (var ($varname))
+ # Add more sections as needed.
+ foreach my $section ('0'..'9', 'n', 'l')
{
- $sections{$section} = 1;
- $varname = '$(' . $varname . ')';
- $vlist{$varname} = 1;
+ my $varname = $pfx . 'man' . $section . '_MANS';
+ if (var ($varname))
+   {
+ $sections{$section} = 1;
+ $varname = '$(' . $varname . ')';
+ $vlist{$varname} = 1;
 
- push_dist_common ($varname)
-   if $pfx eq 'dist_';
+ push_dist_common ($varname)
+   if $pfx eq 'dist_';
+   }
}
-   }
 
-  my $varname = $pfx . 'man_MANS';
-  my $var = var ($varname);
-  if ($var)
-   {
- foreach ($var-value_as_list_recursive)
+ my $varname = $pfx . 'man_MANS';
+ my $var = var ($varname);
+ if ($var)
{
- # A page like `foo.1c' goes into man1dir.
- if (/\.([0-9a-z])([a-z]*)$/)
+ foreach ($var-value_as_list_recursive)
{
- $sections{$1} = 1;
+ # A page like `foo.1c' goes into man1dir.
+ if (/\.([0-9a-z])([a-z]*)$/)
+   {
+ $sections{$1} = 1;
+   }
}
-   }
 
- $varname = '$(' . $varname . ')';
- $vlist{$varname} = 1;
- push_dist_common ($varname)
-   if $pfx eq 'dist_';
+ $varname = '$(' . $varname . ')';
+ $vlist{$varname} = 1;
+ push_dist_common ($varname)
+   if $pfx eq 'dist_';
+   }
}
-}
 
   return unless %sections;
 
diff -ur -N -x 'automake.info*' -x version.texi -x stamp-vti 
automake-1.10.1.orig/automake.in automake-1.10.1/automake.in
--- automake-1.10.1.orig/automake.in2008-02-19 18:07:41.0 +0100
+++ automake-1.10.1/automake.in 2008-02-20 01:52:26.0 +0100
@@ -3363,27 +3363,38 @@
   # Find all the sections in use.  We do this by first looking for
   # standard sections, and then looking for any additional
   # sections used in man_MANS.
-  my (%sections, %vlist);
+  my (%sections, %nsections, %tsections, %nvlist, %tvlist, %nsvlist, %tsvlist);
   # We handle nodist_ for uniformity.  man pages aren't distributed
   # by default so it isn't actually very important.
+  foreach my $npfx ('', 'notrans_')
+{