Patch to allow texi2any for EMACS manual compilation

Patch to allow texi2any for EMACS manual compilation

[Vincent Belaïche]

[-- Attachment #1: Type: text/plain, Size: 815 bytes --]

Hello,

I propose the following patch (attached) so that one can have the
makeinfo program somehow made pointing at texi2any.

motivation:

- makeinfo (C based) is no longer maintained, so sooner or later we will
  have to allow manual compilation with texi2any
  
- texi2any command line is a superset of makeinfo, so there may be
  installations where calling makeinfo actually calls texi2any (or any
  other tricks like setting MAKEINFO to texi2any before calling
  configure --- e.g., at least on my machine, makeinfo.exe is some
  wrapper that calls texi2any under the hood.

- I think that texi2any versioning commences at 5.0 --- maybe Karl or
  Patrice can confirm, so I think that the test on version number in the
  attached patch would be valid.

If people agree I can do the change.

VBR,
  Vincent.




[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: configure_ac.diff --]
[-- Type: text/x-patch, Size: 1115 bytes --]

=== modified file 'configure.ac'
--- configure.ac	2014-09-15 00:20:21 +0000
+++ configure.ac	2014-09-21 20:40:19 +0000
@@ -1102,13 +1102,20 @@
   if test "$MAKEINFO" = "${am_missing_run}makeinfo"; then
     MAKEINFO=makeinfo
   fi
-  case `($MAKEINFO --version) 2>/dev/null` in
-    'makeinfo (GNU texinfo) '4.[[7-9]]* | \
-    'makeinfo (GNU texinfo) '4.[[1-9][0-9]]* | \
-    'makeinfo (GNU texinfo) '[[5-9]]* | \
-    'makeinfo (GNU texinfo) '[[1-9][0-9]]* ) ;;
-    *) MAKEINFO=no;;
-  esac
+  MAKEINFO_VERSION=`($MAKEINFO --version) 2>/dev/null`;
+  declare -i MAKEINFO_CHECK=0;
+  for MAKEINFO_IMPLEMENTATION in makeinfo texi2any; do
+      case $MAKEINFO_VERSION in
+	  "$MAKEINFO_IMPLEMENTATION (GNU texinfo) "4.[7-9]* | \
+          "$MAKEINFO_IMPLEMENTATION (GNU texinfo) "4.[1-9][0-9]* | \
+	  "$MAKEINFO_IMPLEMENTATION (GNU texinfo) "[5-9]* | \
+	  "$MAKEINFO_IMPLEMENTATION (GNU texinfo) "[1-9][0-9]*  )
+	      MAKEINFO_CHECK=$MAKEINFO_CHECK+1;;
+      esac
+  done
+  if [ $MAKEINFO_CHECK -eq 0 ]; then
+      MAKEINFO=no;
+  fi
 fi
 
 ## Makeinfo is unusual.  For a released Emacs, the manuals are

Re: Patch to allow texi2any for EMACS manual compilation

[Karl Berry]

Yes, texi2any 5.0 was its first released version.

k

Re: Patch to allow texi2any for EMACS manual compilation

[Paul Eggert]

Vincent Belaïche wrote:
> I propose the following patch (attached) so that one can have the
> makeinfo program somehow made pointing at texi2any.

I don't see how that patch would make a difference.  Older texinfo 
versions don't have texi2any and 'configure' would continue to use 
'makeinfo' as before.  And in newer texinfo versions, 'texi2any' and 
'makeinfo' are aliases and 'configure' would also continue to use 
'makeinfo' as before.  So why bother to change 'configure'?

RE: Patch to allow texi2any for EMACS manual compilation

[Vincent Belaïche]

[-- Attachment #1: Type: text/plain, Size: 1277 bytes --]

Answers below,

----------------------------------------
> Date: Sun, 21 Sep 2014 14:18:19 -0700
> From: eggert@cs.ucla.edu
> To: vincent.b.1@hotmail.fr; emacs-devel@gnu.org; karl@freefriends.org; pertusus@free.fr
> Subject: Re: Patch to allow texi2any for EMACS manual compilation
>
> Vincent Belaïche wrote:
> > I propose the following patch (attached) so that one can have the
> > makeinfo program somehow made pointing at texi2any.
>
> I don't see how that patch would make a difference. Older texinfo
> versions don't have texi2any and 'configure' would continue to use
> 'makeinfo' as before. And in newer texinfo versions, 'texi2any' and
> 'makeinfo' are aliases and 'configure' would also continue to use
> 'makeinfo' as before. So why bother to change 'configure'?
>

Yes, with the newer texinfo versions configure would also continue to
use makeinfo *unless* the user has configured his machine so that
makeinfo actually calls tex2iany behind the scene.

I think that this would also happen if the user does

MAKEINFO=texi2any ./configure ... <-- with here other args

But I am not sure --- I have not tested it.

  Vincent.

PS: BTW, I did some update to the patch for the sake of simplification
and handling $MAKEINFO = texi2any.


[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: configure.ac.diff --]
[-- Type: text/x-patch, Size: 1375 bytes --]

=== modified file 'configure.ac'
--- configure.ac	2014-09-15 00:20:21 +0000
+++ configure.ac	2014-09-22 17:36:48 +0000
@@ -1099,16 +1099,25 @@
 
 ## Need makeinfo >= 4.7 (?) to build the manuals.
 if test "$MAKEINFO" != "no"; then
-  if test "$MAKEINFO" = "${am_missing_run}makeinfo"; then
-    MAKEINFO=makeinfo
+  for MAKEINFO_IMPLEMENTATION in makeinfo texi2any; do
+      if test "$MAKEINFO" = "${am_missing_run}${MAKEINFO_IMPLEMENTATION}"; then
+         MAKEINFO=${MAKEINFO_IMPLEMENTATION}
+      fi
+  done
+  MAKEINFO_VERSION=`($MAKEINFO --version) 2>/dev/null`;
+  MAKEINFO_CHECK=no;
+  for MAKEINFO_IMPLEMENTATION in makeinfo texi2any; do
+      case $MAKEINFO_VERSION in
+	  "$MAKEINFO_IMPLEMENTATION (GNU texinfo) "4.[7-9]* | \
+          "$MAKEINFO_IMPLEMENTATION (GNU texinfo) "4.[1-9][0-9]* | \
+	  "$MAKEINFO_IMPLEMENTATION (GNU texinfo) "[5-9]* | \
+	  "$MAKEINFO_IMPLEMENTATION (GNU texinfo) "[1-9][0-9]*  )
+	      MAKEINFO_CHECK=yes;;
+      esac
+  done
+  if [ $MAKEINFO_CHECK != yes ]; then
+      MAKEINFO=no;
   fi
-  case `($MAKEINFO --version) 2>/dev/null` in
-    'makeinfo (GNU texinfo) '4.[[7-9]]* | \
-    'makeinfo (GNU texinfo) '4.[[1-9][0-9]]* | \
-    'makeinfo (GNU texinfo) '[[5-9]]* | \
-    'makeinfo (GNU texinfo) '[[1-9][0-9]]* ) ;;
-    *) MAKEINFO=no;;
-  esac
 fi
 
 ## Makeinfo is unusual.  For a released Emacs, the manuals are

Re: Patch to allow texi2any for EMACS manual compilation

[Andreas Schwab]

Vincent Belaïche <vincent.b.1@hotmail.fr> writes:

> === modified file 'configure.ac'
> --- configure.ac	2014-09-15 00:20:21 +0000
> +++ configure.ac	2014-09-22 17:36:48 +0000
> @@ -1099,16 +1099,25 @@
>  
>  ## Need makeinfo >= 4.7 (?) to build the manuals.
>  if test "$MAKEINFO" != "no"; then
> -  if test "$MAKEINFO" = "${am_missing_run}makeinfo"; then
> -    MAKEINFO=makeinfo
> +  for MAKEINFO_IMPLEMENTATION in makeinfo texi2any; do
> +      if test "$MAKEINFO" = "${am_missing_run}${MAKEINFO_IMPLEMENTATION}"; then
> +         MAKEINFO=${MAKEINFO_IMPLEMENTATION}
> +      fi
> +  done
> +  MAKEINFO_VERSION=`($MAKEINFO --version) 2>/dev/null`;
> +  MAKEINFO_CHECK=no;
> +  for MAKEINFO_IMPLEMENTATION in makeinfo texi2any; do
> +      case $MAKEINFO_VERSION in
> +	  "$MAKEINFO_IMPLEMENTATION (GNU texinfo) "4.[7-9]* | \
> +          "$MAKEINFO_IMPLEMENTATION (GNU texinfo) "4.[1-9][0-9]* | \
> +	  "$MAKEINFO_IMPLEMENTATION (GNU texinfo) "[5-9]* | \
> +	  "$MAKEINFO_IMPLEMENTATION (GNU texinfo) "[1-9][0-9]*  )
> +	      MAKEINFO_CHECK=yes;;
> +      esac
> +  done
> +  if [ $MAKEINFO_CHECK != yes ]; then
> +      MAKEINFO=no;
>    fi
> -  case `($MAKEINFO --version) 2>/dev/null` in
> -    'makeinfo (GNU texinfo) '4.[[7-9]]* | \
> -    'makeinfo (GNU texinfo) '4.[[1-9][0-9]]* | \
> -    'makeinfo (GNU texinfo) '[[5-9]]* | \
> -    'makeinfo (GNU texinfo) '[[1-9][0-9]]* ) ;;

Why can't you just replace it with a wildcard?  What's the point in
matching the exact program name?

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

RE: Patch to allow texi2any for EMACS manual compilation

[Vincent Belaïche]

[-- Attachment #1: Type: text/plain, Size: 2962 bytes --]

Answers below...

----------------------------------------
> From: schwab@linux-m68k.org
> To: vincent.b.1@hotmail.fr
> Subject: Re: Patch to allow texi2any for EMACS manual compilation
> Date: Mon, 22 Sep 2014 20:59:48 +0200
> CC: pertusus@free.fr; karl@freefriends.org; eggert@cs.ucla.edu; emacs-devel@gnu.org
>
> Vincent Belaïche <vincent.b.1@hotmail.fr> writes:
>
> > === modified file 'configure.ac'
> > --- configure.ac 2014-09-15 00:20:21 +0000
> > +++ configure.ac 2014-09-22 17:36:48 +0000
> > @@ -1099,16 +1099,25 @@
> >
> > ## Need makeinfo >= 4.7 (?) to build the manuals.
> > if test "$MAKEINFO" != "no"; then
> > - if test "$MAKEINFO" = "${am_missing_run}makeinfo"; then
> > - MAKEINFO=makeinfo
> > + for MAKEINFO_IMPLEMENTATION in makeinfo texi2any; do
> > + if test "$MAKEINFO" = "${am_missing_run}${MAKEINFO_IMPLEMENTATION}"; then
> > + MAKEINFO=${MAKEINFO_IMPLEMENTATION}
> > + fi
> > + done
> > + MAKEINFO_VERSION=`($MAKEINFO --version) 2>/dev/null`;
> > + MAKEINFO_CHECK=no;
> > + for MAKEINFO_IMPLEMENTATION in makeinfo texi2any; do
> > + case $MAKEINFO_VERSION in
> > + "$MAKEINFO_IMPLEMENTATION (GNU texinfo) "4.[7-9]* | \
> > + "$MAKEINFO_IMPLEMENTATION (GNU texinfo) "4.[1-9][0-9]* | \
> > + "$MAKEINFO_IMPLEMENTATION (GNU texinfo) "[5-9]* | \
> > + "$MAKEINFO_IMPLEMENTATION (GNU texinfo) "[1-9][0-9]* )
> > + MAKEINFO_CHECK=yes;;
> > + esac
> > + done
> > + if [ $MAKEINFO_CHECK != yes ]; then
> > + MAKEINFO=no;
> > fi
> > - case `($MAKEINFO --version) 2>/dev/null` in
> > - 'makeinfo (GNU texinfo) '4.[[7-9]]* | \
> > - 'makeinfo (GNU texinfo) '4.[[1-9][0-9]]* | \
> > - 'makeinfo (GNU texinfo) '[[5-9]]* | \
> > - 'makeinfo (GNU texinfo) '[[1-9][0-9]]* ) ;;
>
> Why can't you just replace it with a wildcard? What's the point in
> matching the exact program name?
>

There is a version number checking. So we rely on the fact that texi2any
and makeinfo have both:

- consistent version numbers: C based makeinfo version does not go
  beyond 4.99 (AFAIK last version is 4.13), while texi2any starts at 5.0

- consistent command line, texi2any command line is a superset of that
  makeinfo.

Now, yes it would be possible to use a wildecard, given that it is quite
unlikely than there are other alternative to makeinfo than texi2any.

To make it slightly more restictive, wildcard could be ???????? instead
of * because makeinfo & texi2any both have 8 letters.

BTW, when I did the patch I initially thought that my problem was with
using a fake makeinfo that is indeed a skin of texi2any. But it came out
that this was not the only issue.  Current configure.ac uses that sort
of wildcard:

   [[7-9]]

with [...] encapsulation around `[7-9]'. This does not work for me and I
did not found any such thing in the manual:

http://www.gnu.org/software/bash/manual/bashref.html#pattern-matching

so I removed also that encapsulation as in the attached patch.


[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: configure.ac.diff --]
[-- Type: text/x-patch, Size: 667 bytes --]

=== modified file 'configure.ac'
--- configure.ac	2014-09-15 00:20:21 +0000
+++ configure.ac	2014-09-23 20:42:43 +0000
@@ -1103,10 +1103,11 @@
     MAKEINFO=makeinfo
   fi
   case `($MAKEINFO --version) 2>/dev/null` in
-    'makeinfo (GNU texinfo) '4.[[7-9]]* | \
-    'makeinfo (GNU texinfo) '4.[[1-9][0-9]]* | \
-    'makeinfo (GNU texinfo) '[[5-9]]* | \
-    'makeinfo (GNU texinfo) '[[1-9][0-9]]* ) ;;
+  # ???????? matches both makeinfo & texi2any
+    ????????' (GNU texinfo) '4.[7-9]* | \
+    ????????' (GNU texinfo) '4.[1-9][0-9]* | \
+    ????????' (GNU texinfo) '[5-9]* | \
+    ????????' (GNU texinfo) '[1-9][0-9]* ) ;;
     *) MAKEINFO=no;;
   esac
 fi


[-- Attachment #3: Type: text/plain, Size: 256 bytes --]


If people agree with this patch I can do the change.

VBR,

   Vincent.

> Andreas.
>
> --
> Andreas Schwab, schwab@linux-m68k.org
> GPG Key fingerprint = 58CA 54C7 6D53 942B 1756 01D3 44D5 214B 8276 4ED5
> "And now for something completely different."
>

Re: Patch to allow texi2any for EMACS manual compilation

[Glenn Morris]

Vincent Belaïche wrote:

> BTW, when I did the patch I initially thought that my problem was with
> using a fake makeinfo that is indeed a skin of texi2any.

Please take a step back and tell us clearly what problem you think you
are trying to solve. I've missed that, and so far this is making no sense
to me.

> But it came out that this was not the only issue. Current configure.ac
> uses that sort of wildcard:
>
>    [[7-9]]
>
> with [...] encapsulation around `[7-9]'. This does not work for me and I
> did not found any such thing in the manual:
>
> http://www.gnu.org/software/bash/manual/bashref.html#pattern-matching
>
> so I removed also that encapsulation as in the attached patch.

The double [[]] is necessary for autoconf, which is why you don't find
it in the bash manual. Removing it makes no sense.

Re: Patch to allow texi2any for EMACS manual compilation

[Paul Eggert]

Glenn Morris wrote:
> Please take a step back and tell us clearly what problem you think you
> are trying to solve.

I think he's saying the behavior of Texinfo 5's 'makeinfo --help' 
depends on the name used to invoke makeinfo (which is not following the 
GNU Coding Standards, by the way -- Karl, should I file a Texinfo bug 
report or will this email suffice?), and this means Emacs's 'configure' 
script shouldn't care what that part of the output looks like.  I 
installed a patch along those lines as Emacs trunk bzr 117943.

RE: Patch to allow texi2any for EMACS manual compilation

[Vincent Belaïche]

----------------------------------------
> Date: Wed, 24 Sep 2014 20:05:54 -0700
> From: eggert@cs.ucla.edu
> To: rgm@gnu.org; vincent.b.1@hotmail.fr
> Subject: Re: Patch to allow texi2any for EMACS manual compilation
> CC: karl@freefriends.org; emacs-devel@gnu.org
>
> Glenn Morris wrote:
>> Please take a step back and tell us clearly what problem you think you
>> are trying to solve.
>
Sorry, I have been creating some confusion, because when trying to solve my initial problem (that is configure not accepting that texi2any is used for compiling the documentation) I have come across another one, i.e., that the [...] encapsulation of interval wildcard (e.g. `[7-9]') does not work for me, and I am not even sure that it should work at all because I did not find its documentation in bash manual --- but I suspect that I am just missing the point. Maybe I should open a separate thread for that one ?

> I think he's saying the behavior of Texinfo 5's 'makeinfo --help'
> depends on the name used to invoke makeinfo (which is not following the
> GNU Coding Standards, by the way -- Karl, should I file a Texinfo bug
> report or will this email suffice?), and this means Emacs's 'configure'
> script shouldn't care what that part of the output looks like. I
> installed a patch along those lines as Emacs trunk bzr 117943.
>

I think this is not a bug in Texinfo. texi2any and makeinfo are two different programs, but texi2any can do whatever makeinfo can, and much more. texi2any is thought as a replacement of makeinfo, that is why the version number has some continuity.
   Vincent. 		 	   		  

RE: Patch to allow texi2any for EMACS manual compilation

[Vincent Belaïche]

----------------------------------------
> Date: Wed, 24 Sep 2014 20:05:54 -0700
> From: eggert@cs.ucla.edu
> To: rgm@gnu.org; vincent.b.1@hotmail.fr
> Subject: Re: Patch to allow texi2any for EMACS manual compilation
> CC: karl@freefriends.org; emacs-devel@gnu.org
>
> Glenn Morris wrote:
>> Please take a step back and tell us clearly what problem you think you
>> are trying to solve.
>
Sorry, I have been creating some confusion, because when trying to solve my initial problem (that is configure not accepting that texi2any is used for compiling the documentation) I have come across another one, i.e., that the [...] encapsulation of interval wildcard (e.g. `[7-9]') does not work for me, and I am not even sure that it should work at all because I did not find its documentation in bash manual --- but I suspect that I am just missing the point. Maybe I should open a separate thread for that one ?

> I think he's saying the behavior of Texinfo 5's 'makeinfo --help'
> depends on the name used to invoke makeinfo (which is not following the
> GNU Coding Standards, by the way -- Karl, should I file a Texinfo bug
> report or will this email suffice?), and this means Emacs's 'configure'
> script shouldn't care what that part of the output looks like. I
> installed a patch along those lines as Emacs trunk bzr 117943.
>

I think this is not a bug in Texinfo. texi2any and makeinfo are two different programs, but texi2any can do whatever makeinfo can, and much more. texi2any is thought as a replacement of makeinfo, that is why the version number has some continuity.
   Vincent. 		 	   		  

RE: Patch to allow texi2any for EMACS manual compilation

[Vincent Belaïche]

[-- Attachment #1: Type: text/plain, Size: 492 bytes --]

Ooops... please ignore the first statement about [...] encapsulation in
my previous email, I had missed Glenn's answer that these are simply M4
quotes --- the full story is that in order not to run the full configure
which is really lengthy I did some trials with a simple cut & paste
without paying attention to dequoting.

So in a nutshell, I meant that I do not think that this is a bug in
Texinfo, because texi2any and makeinfo are two different programs.

   Vincent.

PS: patch update.

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: configure.ac.diff --]
[-- Type: text/x-patch, Size: 675 bytes --]

=== modified file 'configure.ac'
--- configure.ac	2014-09-15 00:20:21 +0000
+++ configure.ac	2014-09-25 04:41:34 +0000
@@ -1103,10 +1103,11 @@
     MAKEINFO=makeinfo
   fi
   case `($MAKEINFO --version) 2>/dev/null` in
-    'makeinfo (GNU texinfo) '4.[[7-9]]* | \
-    'makeinfo (GNU texinfo) '4.[[1-9][0-9]]* | \
-    'makeinfo (GNU texinfo) '[[5-9]]* | \
-    'makeinfo (GNU texinfo) '[[1-9][0-9]]* ) ;;
+  # ???????? matches both makeinfo & texi2any
+    ????????' (GNU texinfo) '4.[[7-9]]* | \
+    ????????' (GNU texinfo) '4.[[1-9][0-9]]* | \
+    ????????' (GNU texinfo) '[[5-9]]* | \
+    ????????' (GNU texinfo) '[[1-9][0-9]]* ) ;;
     *) MAKEINFO=no;;
   esac
 fi


[-- Attachment #3: Type: text/plain, Size: 1905 bytes --]



----------------------------------------
> From: vincent.b.1@hotmail.fr
> To: eggert@cs.ucla.edu; rgm@gnu.org
> CC: karl@freefriends.org; emacs-devel@gnu.org
> Subject: RE: Patch to allow texi2any for EMACS manual compilation
> Date: Thu, 25 Sep 2014 06:33:34 +0200
>
>
>
> ----------------------------------------
> > Date: Wed, 24 Sep 2014 20:05:54 -0700
> > From: eggert@cs.ucla.edu
> > To: rgm@gnu.org; vincent.b.1@hotmail.fr
> > Subject: Re: Patch to allow texi2any for EMACS manual compilation
> > CC: karl@freefriends.org; emacs-devel@gnu.org
> >
> > Glenn Morris wrote:
> >> Please take a step back and tell us clearly what problem you think you
> >> are trying to solve.
> >
> Sorry, I have been creating some confusion, because when trying to solve my initial problem (that is configure not accepting that texi2any is used for compiling the documentation) I have come across another one, i.e., that the [...] encapsulation of interval wildcard (e.g. `[7-9]') does not work for me, and I am not even sure that it should work at all because I did not find its documentation in bash manual --- but I suspect that I am just missing the point. Maybe I should open a separate thread for that one ?
>
> > I think he's saying the behavior of Texinfo 5's 'makeinfo --help'
> > depends on the name used to invoke makeinfo (which is not following the
> > GNU Coding Standards, by the way -- Karl, should I file a Texinfo bug
> > report or will this email suffice?), and this means Emacs's 'configure'
> > script shouldn't care what that part of the output looks like. I
> > installed a patch along those lines as Emacs trunk bzr 117943.
> >
>
> I think this is not a bug in Texinfo. texi2any and makeinfo are two different programs, but texi2any can do whatever makeinfo can, and much more. texi2any is thought as a replacement of makeinfo, that is why the version number has some continuity.
> Vincent.

RE: Patch to allow texi2any for EMACS manual compilation

[Vincent Belaïche]

I saw that Paul did the change which I expected (with a * wildcard),
that closes the point for me.

Thank you.

   Vincent.