Bootstrapping Unicode branch on Windows

Bootstrapping Unicode branch on Windows

[Juanma Barranquero]

Does the emacs-unicode-2 branch bootstrap on Windows? I'm getting this error:

 [...]
 Generating autoloads for kermit.el...done
 Generating autoloads for kmacro.el...
 Generating autoloads for kmacro.el...done
 Local variables entry is missing the suffix
 make[1]: *** [autoloads] Error -1
 make[1]: Leaving directory `C:/bin/emacs/unicode/lisp'
 make: *** [bootstrap-gmake] Error 2

                    /L/e/k/t/u

Re: Bootstrapping Unicode branch on Windows

[Juanma Barranquero]

On 2006-12-04, I wrote:

> Does the emacs-unicode-2 branch bootstrap on Windows? I'm getting this error:
>
>  [...]
>  Generating autoloads for kermit.el...done
>  Generating autoloads for kmacro.el...
>  Generating autoloads for kmacro.el...done
>  Local variables entry is missing the suffix
>  make[1]: *** [autoloads] Error -1
>  make[1]: Leaving directory `C:/bin/emacs/unicode/lisp'
>  make: *** [bootstrap-gmake] Error 2

(I forgot to say that is with Windows XP, MinGW 3.4.5, and using CMD, not sh.)

Well, it seems the problem is related to Emacs believing that some ^M
at the end of the lines are part of the local variables suffix.

I've been able to get much farther down the bootstrapping path by:

 - checking out lisp/international/uni-*.el with -kb (which "cures"
the problem above)
 - setting LANG=en_US.UTF-8 (this avoids a lot of warnings saying that
iso-latin-1 disagrees with system codeset cp1252)
 - checking out assorted quail/*/* files with -kb (though I'm not
entirely sure that's needed)

However, unicode Emacs refuses to compile leim/quail/PY.el;
apparently, it has issues reading it:

In toplevel form:
quail/PY.el:87:1:Error: Invalid read syntax: "#"

Any idea?
                    /L/e/k/t/u

Re: Bootstrapping Unicode branch on Windows

[Jason Rumney]

Juanma Barranquero wrote:
> Well, it seems the problem is related to Emacs believing that some ^M
> at the end of the lines are part of the local variables suffix.

I tried checking out and bootstrapping emacs-unicode-2 a couple of days ago.

I had to check out the following files with -kb:

nt/* (this is the case with CVS HEAD also)

lisp/ldefs-boot.el
lisp/international/uni-*.el
leim/CXTERM-DIC/*
leim/MISC-DIC/pinyin.map

I think it was this last one that affected compilation of PY.el

Re: Bootstrapping Unicode branch on Windows

[Juanma Barranquero]

On 2/8/07, Jason Rumney <jasonr@gnu.org> wrote:

> nt/* (this is the case with CVS HEAD also)
> lisp/ldefs-boot.el
> lisp/international/uni-*.el
> leim/CXTERM-DIC/*
> leim/MISC-DIC/pinyin.map

All these I have with -kb (I just re-checked it).

> I think it was this last one that affected compilation of PY.el

There's something else at work.

What size is your leim/quail/PY.el? (I'm trying to guess if it's wrong
because of some coding issue or something like that). Mine is 25.130
bytes, and it's a CRLF file.

                    /L/e/k/t/u

Re: Bootstrapping Unicode branch on Windows

[Zhang Wei]

"Juanma Barranquero" <lekktu@gmail.com> writes:

[...]

> Well, it seems the problem is related to Emacs believing that some ^M
> at the end of the lines are part of the local variables suffix.

Try to checkout the source code with -kb option:

cvs -z3 -d:pserver:anonymous@cvs.savannah.gnu.org:/sources/emacs \
    co -kb -r emacs-unicode-2 -d emacs-unicode-2 emacs

otherwise it won't compile on Windows.

Re: Bootstrapping Unicode branch on Windows

[Juanma Barranquero]

On 2/8/07, Zhang Wei <id.brep@gmail.com> wrote:

> Try to checkout the source code with -kb option:

Thanks for your suggestion.

I check out several files with -kb (lisp/international/uni-*.el, nt/*,
leim/quail/MISC-DIC/*). I don't think it's really necessary to check
out the full source tree with -kb.

I'm not yet sure what my problem is, but the symptom is Emacs
generating a wrong leim/quail/PY.el file. I'll have to dig a little
more.

                    /L/e/k/t/u

Re: Bootstrapping Unicode branch on Windows

[Juanma Barranquero]

Finally I've been able to solve the mystery, and it's my fault...

While trying to work around the initial issue (with
lisp/international/uni-*.el files) I did several tests with and
without -kb. In one of these tests, I accidentally updated most of
leim/*/* with

  cvs -q update -C -A

instead of

  cvs -q update -C -A -r emacs-unicode-2

so I had an emacs-unicode-2 source tree with leim/* switched back to
HEAD. Definitely Not Recommendable.

All in all, I'd suggest adding a note somewhere in the emacs-unicode-2
branch (nt/INSTALL, perhaps) clarifying which files need to be checked
out with -kb, and the trick of defining a LANG environment variable
with UTF-8 to avoid warnings about conflicting codesets. This will
make life easier for anyone trying to compile the unicode branch (and
soon the HEAD, hopefully) on Windows.

                    /L/e/k/t/u

Re: Bootstrapping Unicode branch on Windows

[Eli Zaretskii]

> Date: Fri, 9 Feb 2007 11:27:26 +0100
> From: "Juanma Barranquero" <lekktu@gmail.com>
> 
> All in all, I'd suggest adding a note somewhere in the emacs-unicode-2
> branch (nt/INSTALL, perhaps) clarifying which files need to be checked
> out with -kb, and the trick of defining a LANG environment variable
> with UTF-8 to avoid warnings about conflicting codesets.

Actually, I think these are all signs of bugs that should be fixed.

Re: Bootstrapping Unicode branch on Windows

[Juanma Barranquero]

On 2/9/07, Eli Zaretskii <eliz@gnu.org> wrote:

> Actually, I think these are all signs of bugs that should be fixed.

The CRLF vs LF vs CRCRLF issues, for sure, after the unicode-2 merge.

I was talking of saving some poor soul the time I've spent fighting
these hidden issues. But I guess isn't worth the time if we release
soon (hey, a guy can hope...) and merge the unicode-2 branch.

                    /L/e/k/t/u

Re: Bootstrapping Unicode branch on Windows

[Kenichi Handa]

In article <f7ccd24b0702081731j281b042ay1759e484db0a33c6@mail.gmail.com>, "Juanma Barranquero" <lekktu@gmail.com> writes:

> On 2/8/07, Zhang Wei <id.brep@gmail.com> wrote:
> > Try to checkout the source code with -kb option:

> Thanks for your suggestion.

> I check out several files with -kb (lisp/international/uni-*.el, nt/*,
> leim/quail/MISC-DIC/*). I don't think it's really necessary to check
> out the full source tree with -kb.

> I'm not yet sure what my problem is, but the symptom is Emacs
> generating a wrong leim/quail/PY.el file. I'll have to dig a little
> more.

I've just fixed titdic-cnv.el so that it always detects EOL
format on reading files and it always write files with
unix-like EOL format.  So, now CXTERM-DIC/* and MISC-IDC/*
can be checked out both by LF and by CRLF.

The change is done in HEAD, so it may take some time that
to be propagated to emacs-unicode-2.

---
Kenichi Handa
handa@m17n.org

Re: Bootstrapping Unicode branch on Windows

[Juanma Barranquero]

On 2/11/07, Kenichi Handa <handa@m17n.org> wrote:

> So, now CXTERM-DIC/* and MISC-IDC/*
> can be checked out both by LF and by CRLF.

Thanks.

Note however that leim/MISC-DIC/cangjie-table.* (in both the HEAD
branch and emacs-unicode-2) do contain CRLF and are not marked as
binary.

                    /L/e/k/t/u

Re: Bootstrapping Unicode branch on Windows

[Kenichi Handa]

In article <f7ccd24b0702120141j5b9ae8c6i9e23f1451f259378@mail.gmail.com>, "Juanma Barranquero" <lekktu@gmail.com> writes:

> On 2/11/07, Kenichi Handa <handa@m17n.org> wrote:
> > So, now CXTERM-DIC/* and MISC-IDC/*
> > can be checked out both by LF and by CRLF.

> Thanks.

> Note however that leim/MISC-DIC/cangjie-table.* (in both the HEAD
> branch and emacs-unicode-2) do contain CRLF and are not marked as
> binary.

They contain CRLF originally.  So I dared not change it.  It
seems this change (for tsang-quick-converter) by Jason
already takes care of such a case.

      ;; Handle double CR line ends, which result when checking out of
      ;; CVS on MS-Windows.
      (goto-char (point-min))
      (while (re-search-forward "\r\r$" nil t)
	(replace-match ""))

---
Kenichi Handa
handa@m17n.org

Re: Bootstrapping Unicode branch on Windows

[Juanma Barranquero]

On 2/12/07, Kenichi Handa <handa@m17n.org> wrote:

> They contain CRLF originally.  So I dared not change it.

Well, not now obviously, but after the release we should try to fix
them in the repository and see whether something breaks.

>       ;; Handle double CR line ends, which result when checking out of
>       ;; CVS on MS-Windows.

Aha, I see.

Thanks,
                    Juanma

Re: Bootstrapping Unicode branch on Windows

[Jason Rumney]

Juanma Barranquero wrote:
> On 2/12/07, Kenichi Handa <handa@m17n.org> wrote:
>
>> They contain CRLF originally.  So I dared not change it.
>
> Well, not now obviously, but after the release we should try to fix
> them in the repository and see whether something breaks.
I don't think its a case of something breaking if they are changed, more 
that changing them compared to the original source makes updates more 
complex and error prone.

Re: Bootstrapping Unicode branch on Windows

[Juanma Barranquero]

On 2/12/07, Jason Rumney <jasonr@gnu.org> wrote:

> I don't think its a case of something breaking if they are changed, more
> that changing them compared to the original source makes updates more
> complex and error prone.

I'm not sure I understand. Do you mean that the original files contain
CRLF? In that case, perhaps they should be checked in with -kb; these
are not text files in the CVS meaning of the word.

             Juanma

Re: Bootstrapping Unicode branch on Windows

[Eli Zaretskii]

> Date: Mon, 12 Feb 2007 16:13:18 +0100
> From: "Juanma Barranquero" <lekktu@gmail.com>
> Cc: id.brep@gmail.com, emacs-devel@gnu.org, Kenichi Handa <handa@m17n.org>
> 
> I'm not sure I understand. Do you mean that the original files contain
> CRLF? In that case, perhaps they should be checked in with -kb; these
> are not text files in the CVS meaning of the word.

Using -kb has disadvantages: it causes CVS to be unable to send
patches when you update from the repository; instead, it sends the
entire file downstream.  With huge files such as the ones we are
discussing, that would be a nuisance.

Re: Bootstrapping Unicode branch on Windows

[Juanma Barranquero]

On 2/12/07, Eli Zaretskii <eliz@gnu.org> wrote:

> Using -kb has disadvantages: it causes CVS to be unable to send
> patches when you update from the repository; instead, it sends the
> entire file downstream.

Seems like this one is for the day (if any) we switch to a more modern
version control software.

             Juanma

Re: Bootstrapping Unicode branch on Windows

[Juanma Barranquero]

On 2/11/07, Kenichi Handa <handa@m17n.org> wrote:

> I've just fixed titdic-cnv.el so that it always detects EOL
> format on reading files and it always write files with
> unix-like EOL format.  So, now CXTERM-DIC/* and MISC-IDC/*
> can be checked out both by LF and by CRLF.

With the current HEAD, no CXTERM-DIC/* file needs to be checked out with -kb.

However, with the current emacs-unicode-2 branch, these files still need -kb:

CXTERM-DIC/ETZY.tit
CXTERM-DIC/PY-b5.tit
CXTERM-DIC/TONEPY.tit
CXTERM-DIC/ZOZY.tit

             Juanma