Re: make distcheck not working
[EMAIL PROTECTED] (Lars Gullik Bjønnes) writes:
| [EMAIL PROTECTED] (Lars Gullik Bjønnes) writes:
>
| | make distcheck is not working.
>
| I am getting there, slowly.
YES!
Finally!
Several small fixes... all taking ages to test.
Index: mkinstalldirs
===
RCS file: mkinstalldirs
diff -N mkinstalldirs
--- mkinstalldirs 21 Nov 2000 16:35:36 - 1.1
+++ /dev/null 1 Jan 1970 00:00:00 -
@@ -1,32 +0,0 @@
-#! /bin/sh
-# mkinstalldirs --- make directory hierarchy
-# Author: Noah Friedman <[EMAIL PROTECTED]>
-# Created: 1993-05-16
-# Public domain
-
-errstatus=0
-
-for file
-do
- set fnord `echo ":$file" | sed -ne 's/^:\//#/;s/^://;s/\// /g;s/^#/\//;p'`
- shift
-
- pathcomp=
- for d in ${1+"$@"} ; do
- pathcomp="$pathcomp$d"
- case "$pathcomp" in
- -* ) pathcomp=./$pathcomp ;;
- esac
-
- if test ! -d "$pathcomp"; then
-echo "mkdir $pathcomp" 1>&2
-mkdir "$pathcomp" || errstatus=$?
- fi
-
- pathcomp="$pathcomp/"
- done
-done
-
-exit $errstatus
-
-# mkinstalldirs ends here
Index: config/depcomp.diff
===
RCS file: config/depcomp.diff
diff -N config/depcomp.diff
--- config/depcomp.diff 22 Mar 2002 12:32:11 - 1.1
+++ /dev/null 1 Jan 1970 00:00:00 -
@@ -1,39 +0,0 @@
-Index: depcomp
-===
-RCS file: /usr/local/lyx/cvsroot/lyx-devel/config/depcomp,v
-retrieving revision 1.1
-diff -u -p -r1.1 depcomp
depcomp 17 Oct 2001 15:38:58 - 1.1
-+++ depcomp 22 Mar 2002 12:18:23 -
-@@ -198,14 +198,28 @@ aix)
- ;;
-
- tru64)
-- # The Tru64 AIX compiler uses -MD to generate dependencies as a side
-+ # The Tru64 DEC compiler uses -MD to generate dependencies as a side
-# effect. `cc -MD -o foo.o ...' puts the dependencies into `foo.o.d'.
-# At least on Alpha/Redhat 6.1, Compaq CCC V6.2-504 seems to put
-# dependencies in `foo.d' instead, so we check for that too.
-# Subdirectories are respected.
-
-- tmpdepfile1="$object.d"
-- tmpdepfile2=`echo "$object" | sed -e 's/.o$/.d/'`
-+ # This is incorrect as it assumes that $object has a .o extension when
-+ # it could well have a .lo one.
-+ #tmpdepfile1="$object.d"
-+ # This is just bad sed. The /.o$/ should be /\.o/, nonwithstanding the
-+ # flawed logic, described above.
-+ #tmpdepfile2=`echo "$object" | sed -e 's/.o$/.d/'`
-+
-+ # The correct way:
-+ # Strip the (.o, or .lo) extension from $object
-+ # (Same piece of sed magic as used for the AIX compiler above.)
-+ stripped=`echo "$object" | sed -e 's,^.*/,,' -e 's/\(.*\)\..*$/\1/'`
-+ # Tru64 compiler dependency file.
-+ tmpdepfile1="$stripped.o.d"
-+ # Compaq CCC V6.2 dependency file.
-+ tmpdepfile2="$stripped.d"
-+
-if test "$libtool" = yes; then
- "$@" -Wc,-MD
-else
Index: config/gettext.m4.patch
===
RCS file: config/gettext.m4.patch
diff -N config/gettext.m4.patch
--- config/gettext.m4.patch 8 Jan 2000 21:06:59 - 1.1
+++ /dev/null 1 Jan 1970 00:00:00 -
@@ -1,16 +0,0 @@
/usr/share/aclocal/gettext.m4 Thu Sep 23 05:05:24 1999
-+++ gettext.m4 Sat Jan 8 04:12:49 2000
-@@ -227,7 +227,12 @@
- else
-AC_MSG_CHECKING(for catalogs to be installed)
-NEW_LINGUAS=
-- for lang in ${LINGUAS=$ALL_LINGUAS}; do
-+dnl == Fix is here! ==
-+dnl Some sh do not like substitution in bounds of for loops
-+dnlfor lang in ${LINGUAS=$ALL_LINGUAS}; do
-+tmplinguas=${LINGUAS=$ALL_LINGUAS}
-+for lang in ${tmplinguas}; do
-+dnl end of fix
- case "$ALL_LINGUAS" in
- *$lang*) NEW_LINGUAS="$NEW_LINGUAS $lang" ;;
- esac
Index: lib/Makefile.am
===
RCS file: /usr/local/lyx/cvsroot/lyx-devel/lib/Makefile.am,v
retrieving revision 1.53
diff -u -p -b -r1.53 Makefile.am
--- lib/Makefile.am 1 Apr 2004 08:09:14 - 1.53
+++ lib/Makefile.am 28 May 2004 06:44:33 -
@@ -4,38 +4,910 @@ DISTCLEANFILES = *.orig *.rej *~ *.bak c
MAINTAINERCLEANFILES = $(srcdir)/Makefile.in
SUBDIRS = reLyX
-CHMOD=chmod
-
-BIND = bind/*.bind
-CLIPART = clipart/*.eps
-DOCDIST = doc/LyXConfig.lyx*
-DOCINST = doc/*.lyx doc/*.eps doc/*.lyx.in
-EXAMPLES = examples/*.lyx examples/*.fen
-IMAGES = images/*.xpm images/*.ppm images/math/*.xpm
-KBD = kbd/*.kmap kbd/*.cdef
-LAYOUT = layouts/*.layout layouts/*.inc
-LYXSCRIPTS = configure configure.cmd scripts/* lyx2lyx/*
-TEMPL = templates/*.lyx
-TEXSUPPORT = tex/*.cls tex/*.sty
-UI = ui/*.ui
-
-LIBINSTFILES = $(BIND) $(CLIPART) $(DOCINST) $(EXAMPLES) $(IMAGES) \
- $(KBD) $(LAYOUT) $(TEMPL) $(TEXSUPPORT) $(UI)
+EXTRA_DIST = CREDITS chkconfig.ltx configure.cmd lyxrc.example \
+ external_templates bind clipart doc/LaTeXConfig.lyx.in doc/LyXConfig.lyx.in examples images \
+ kb
Re: make distcheck not working
Angus Leeming <[EMAIL PROTECTED]> writes: | lyx2lyx | already has one by the way. Does it? -- Lgb
Re: make distcheck not working
Angus Leeming <[EMAIL PROTECTED]> writes: | Lars Gullik Bjønnes wrote: > >> [EMAIL PROTECTED] (Lars Gullik Bjønnes) writes: >> >> | make distcheck is not working. >> >> I am getting there, slowly. Doing a "make distcheck" takes time... >> and for a lot of the distcheck stuff you have to run it to >> completeion and from the beginning for everything you fix/test. >> >> Anyway, it seems most of my problems are solved now. Albeit I have >> only tested for xforms. >> >> This are the changes I have done to my tree so far: > | Yuch. What a pain. Ok, some of the changes in lib/Makefile.am is not strictly needed, but it is the way to do it with automake, and not fiddle with automake internals. | Given that lib/Makefile.am is now so big, I think | it makes more sense to have per-subdirectory Makefile.am s. lyx2lyx | already has one by the way. The lyx2lyx_DATA etc should certainly be | moved in there. Hmm... why is the install/dist done from parent dir then? Hmm da hmm... | Index: src/frontends/xforms/forms/Makefile.am | -$(SHELL) $(SCRIPT) $< | +$(LN_S) $< . | +$(SHELL) $(SCRIPT) `basename $<` > | Won't this fail (ln -s) if build dir == src dir? sure the ln -s will fail, but the rest will not. But sure it is not nice... so I should add a check for srcdir != builddir. -- Lgb
Re: make distcheck not working
Lars Gullik Bjønnes wrote: > [EMAIL PROTECTED] (Lars Gullik Bjønnes) writes: > > | make distcheck is not working. > > I am getting there, slowly. Doing a "make distcheck" takes time... > and for a lot of the distcheck stuff you have to run it to > completeion and from the beginning for everything you fix/test. > > Anyway, it seems most of my problems are solved now. Albeit I have > only tested for xforms. > > This are the changes I have done to my tree so far: Yuch. What a pain. Given that lib/Makefile.am is now so big, I think it makes more sense to have per-subdirectory Makefile.am s. lyx2lyx already has one by the way. The lyx2lyx_DATA etc should certainly be moved in there. Index: src/frontends/xforms/forms/Makefile.am -$(SHELL) $(SCRIPT) $< +$(LN_S) $< . +$(SHELL) $(SCRIPT) `basename $<` Won't this fail (ln -s) if build dir == src dir? -- Angus
Re: doc in lyx-devel or out or in ...
Angus Leeming <[EMAIL PROTECTED]> writes: | Lars Gullik Bjønnes wrote: >> I know I brought this up the last time... but what was the >> conclusion of the discussion? >> >> To move do inside lyx-devel? Or to keep it as a separate module that >> is not disted with lyx-x.tar.gz? >> >> The motivation is to get rid of the cvs commands done when doing a >> "make dist" >> > | Given that docs is half the size of po (which I guess is distributed | with the generated files)... > | $ du -ks docs devel; for dir in devel/*; do test -d $dir && du -ks | $dir; done [disk stats snipped] So what you are saying is: "Sure why not put docs in lyx-devel, won't make much of a difference anyway."? I do not think I will do any cvs magic to copy the files over, just make a doc dir (we have it already) and copy the doc files there. We will loose history for those files, but otoh we will still have lyx-doc. (and we should never use HEAD there again only the existing release tags/branches). Does anyone oppose this move? | However, don't you think that it would be nice to distribute | per-language rpms containing just the po file and any translated | docs? Sure, and too would be easier if we had the documentation _in_ lyx-devel | Given that there should really be a en.po now that JMarc has | ASCIIfied the translatable strings, it would even be consistent to | have a lyx_en.rpm lyx-lang-en.rpm would be my preference. -- Lgb
Re: doc in lyx-devel or out or in ...
Lars Gullik Bjønnes wrote: > I know I brought this up the last time... but what was the > conclusion of the discussion? > > To move do inside lyx-devel? Or to keep it as a separate module that > is not disted with lyx-x.tar.gz? > > The motivation is to get rid of the cvs commands done when doing a > "make dist" > Given that docs is half the size of po (which I guess is distributed with the generated files)... $ du -ks docs devel; for dir in devel/*; do test -d $dir && du -ks $dir; done 10820 docs 53716 devel 6508devel/boost 676 devel/config 20 devel/CVS 540 devel/development 568 devel/images 500 devel/intl 7124devel/lib 212 devel/m4 488 devel/patches 20568 devel/po 76 devel/sourcedoc 12684 devel/src However, don't you think that it would be nice to distribute per-language rpms containing just the po file and any translated docs? Given that there should really be a en.po now that JMarc has ASCIIfied the translatable strings, it would even be consistent to have a lyx_en.rpm -- Angus
Re: make distcheck not working
[EMAIL PROTECTED] (Lars Gullik Bjønnes) writes:
| make distcheck is not working.
I am getting there, slowly. Doing a "make distcheck" takes time... and
for a lot of the distcheck stuff you have to run it to completeion and
from the beginning for everything you fix/test.
Anyway, it seems most of my problems are solved now. Albeit I have
only tested for xforms.
This are the changes I have done to my tree so far:
Index: mkinstalldirs
===
RCS file: mkinstalldirs
diff -N mkinstalldirs
--- mkinstalldirs 21 Nov 2000 16:35:36 - 1.1
+++ /dev/null 1 Jan 1970 00:00:00 -
@@ -1,32 +0,0 @@
-#! /bin/sh
-# mkinstalldirs --- make directory hierarchy
-# Author: Noah Friedman <[EMAIL PROTECTED]>
-# Created: 1993-05-16
-# Public domain
-
-errstatus=0
-
-for file
-do
- set fnord `echo ":$file" | sed -ne 's/^:\//#/;s/^://;s/\// /g;s/^#/\//;p'`
- shift
-
- pathcomp=
- for d in ${1+"$@"} ; do
- pathcomp="$pathcomp$d"
- case "$pathcomp" in
- -* ) pathcomp=./$pathcomp ;;
- esac
-
- if test ! -d "$pathcomp"; then
-echo "mkdir $pathcomp" 1>&2
-mkdir "$pathcomp" || errstatus=$?
- fi
-
- pathcomp="$pathcomp/"
- done
-done
-
-exit $errstatus
-
-# mkinstalldirs ends here
Index: config/depcomp.diff
===
RCS file: config/depcomp.diff
diff -N config/depcomp.diff
--- config/depcomp.diff 22 Mar 2002 12:32:11 - 1.1
+++ /dev/null 1 Jan 1970 00:00:00 -
@@ -1,39 +0,0 @@
-Index: depcomp
-===
-RCS file: /usr/local/lyx/cvsroot/lyx-devel/config/depcomp,v
-retrieving revision 1.1
-diff -u -p -r1.1 depcomp
depcomp 17 Oct 2001 15:38:58 - 1.1
-+++ depcomp 22 Mar 2002 12:18:23 -
-@@ -198,14 +198,28 @@ aix)
- ;;
-
- tru64)
-- # The Tru64 AIX compiler uses -MD to generate dependencies as a side
-+ # The Tru64 DEC compiler uses -MD to generate dependencies as a side
-# effect. `cc -MD -o foo.o ...' puts the dependencies into `foo.o.d'.
-# At least on Alpha/Redhat 6.1, Compaq CCC V6.2-504 seems to put
-# dependencies in `foo.d' instead, so we check for that too.
-# Subdirectories are respected.
-
-- tmpdepfile1="$object.d"
-- tmpdepfile2=`echo "$object" | sed -e 's/.o$/.d/'`
-+ # This is incorrect as it assumes that $object has a .o extension when
-+ # it could well have a .lo one.
-+ #tmpdepfile1="$object.d"
-+ # This is just bad sed. The /.o$/ should be /\.o/, nonwithstanding the
-+ # flawed logic, described above.
-+ #tmpdepfile2=`echo "$object" | sed -e 's/.o$/.d/'`
-+
-+ # The correct way:
-+ # Strip the (.o, or .lo) extension from $object
-+ # (Same piece of sed magic as used for the AIX compiler above.)
-+ stripped=`echo "$object" | sed -e 's,^.*/,,' -e 's/\(.*\)\..*$/\1/'`
-+ # Tru64 compiler dependency file.
-+ tmpdepfile1="$stripped.o.d"
-+ # Compaq CCC V6.2 dependency file.
-+ tmpdepfile2="$stripped.d"
-+
-if test "$libtool" = yes; then
- "$@" -Wc,-MD
-else
Index: config/gettext.m4.patch
===
RCS file: config/gettext.m4.patch
diff -N config/gettext.m4.patch
--- config/gettext.m4.patch 8 Jan 2000 21:06:59 - 1.1
+++ /dev/null 1 Jan 1970 00:00:00 -
@@ -1,16 +0,0 @@
/usr/share/aclocal/gettext.m4 Thu Sep 23 05:05:24 1999
-+++ gettext.m4 Sat Jan 8 04:12:49 2000
-@@ -227,7 +227,12 @@
- else
-AC_MSG_CHECKING(for catalogs to be installed)
-NEW_LINGUAS=
-- for lang in ${LINGUAS=$ALL_LINGUAS}; do
-+dnl == Fix is here! ==
-+dnl Some sh do not like substitution in bounds of for loops
-+dnlfor lang in ${LINGUAS=$ALL_LINGUAS}; do
-+tmplinguas=${LINGUAS=$ALL_LINGUAS}
-+for lang in ${tmplinguas}; do
-+dnl end of fix
- case "$ALL_LINGUAS" in
- *$lang*) NEW_LINGUAS="$NEW_LINGUAS $lang" ;;
- esac
Index: lib/Makefile.am
===
RCS file: /usr/local/lyx/cvsroot/lyx-devel/lib/Makefile.am,v
retrieving revision 1.53
diff -u -p -r1.53 Makefile.am
--- lib/Makefile.am 1 Apr 2004 08:09:14 - 1.53
+++ lib/Makefile.am 27 May 2004 16:56:50 -
@@ -4,38 +4,910 @@ DISTCLEANFILES = *.orig *.rej *~ *.bak c
MAINTAINERCLEANFILES = $(srcdir)/Makefile.in
SUBDIRS = reLyX
-CHMOD=chmod
-
-BIND = bind/*.bind
-CLIPART = clipart/*.eps
-DOCDIST = doc/LyXConfig.lyx*
-DOCINST = doc/*.lyx doc/*.eps doc/*.lyx.in
-EXAMPLES = examples/*.lyx examples/*.fen
-IMAGES = images/*.xpm images/*.ppm images/math/*.xpm
-KBD = kbd/*.kmap kbd/*.cdef
-LAYOUT = layouts/*.layout layouts/*.inc
-LYXSCRIPTS = configure configure.cmd scripts/* lyx2lyx/*
-TEMPL = templates/*.lyx
-TEXSUPPORT = tex/*.cls tex/*.sty
-UI = ui/*.ui
-
-LIBINSTFILES = $(BIND) $(CLIPART) $(DOCINST) $(EXAMPLES) $(IMAGES) \
- $(KBD) $(LAYOUT) $(TEMPL) $(TE
doc in lyx-devel or out or in ...
I know I brought this up the last time... but what was the conclusion of the discussion? To move do inside lyx-devel? Or to keep it as a separate module that is not disted with lyx-x.tar.gz? The motivation is to get rid of the cvs commands done when doing a "make dist" -- Lgb
Re: fdesign -convert is buggy
Lars Gullik Bjønnes wrote:
> | } else
> | strcpy(fname, filename);
> | }
> | strcat(fname, ext);
> | }
>
> It almost seems to me that the else clause should just be
>
>strcpy(fname, filename_only(filename));
I think you're correct. Thanks.
> | build_fname(fname, "foo/bar/baz", ".h");
>
> If my above method was used this would produce just baz.h. but the
> old way would produce foo/bar/baz.h
Correct.
The test program seems to work exactly as you want.
Jean-Marc, are you happy with the change?
Index: fdesign/fd_printC.c
===
RCS file: /cvsroot/xforms/xforms/fdesign/fd_printC.c,v
retrieving revision 1.8
diff -u -p -r1.8 fd_printC.c
--- fdesign/fd_printC.c 27 Nov 2003 18:15:00 - 1.8
+++ fdesign/fd_printC.c 27 May 2004 16:50:46 -
@@ -82,7 +82,7 @@ build_fname(char * fname, char const * f
strcat(fname, "/");
strcat(fname, filename_only(filename));
} else
- strcpy(fname, filename);
+ strcpy(fname, filename_only(filename));
strcat(fname, ext);
}
--
Angus#include
#include
#include
typedef struct {
char * output_dir;
} FD_Opt;
FD_Opt fdopt;
static char const *
filename_only(char const * filename)
{
char const * ptr = strrchr(filename, '/');
if (ptr)
return ptr+1;
return filename;
}
/* Given 'filename' (no extension) and 'ext', generate 'fname' */
static void
build_fname(char * fname, char const * filename, char const * ext)
{
if (fdopt.output_dir) {
strcpy(fname, fdopt.output_dir);
if (fname[strlen(fdopt.output_dir) - 1] != '/')
strcat(fname, "/");
strcat(fname, filename_only(filename));
} else
strcpy(fname, filename_only(filename));
strcat(fname, ext);
}
int main(int argc, char * argv[])
{
char fname[PATH_MAX + 1];
fdopt.output_dir = 0;
if (argc > 1) {
fdopt.output_dir = (char *)malloc(strlen(argv[1]) + 1);
strcpy(fdopt.output_dir, argv[1]);
}
build_fname(fname, "foo/bar/baz", ".h");
printf("Output fname is \"%s\"\n", fname);
return 0;
}
Re: fdesign -convert is buggy
Angus Leeming <[EMAIL PROTECTED]> writes:
| I believe that we achieve what you're looking for with
| -} else
| - strcpy(fname, filename);
| +} else {
| + size_t const buffer_size = 512;
| + char buffer[buffer_size];
| + char * cwd = getcwd(buffer, buffer_size);
| + if (cwd) {
| + strcpy(fname, cwd);
| + strcat(fname, "/");
| + strcat(fname, filename_only(filename));
| + } else
| + strcpy(fname, filename);
Why isn't filename_only used here?
| static char const *
| filename_only(char const * filename)
| {
| char const * ptr = strrchr(filename, '/');
| if (ptr)
| return ptr+1;
| return filename;
| }
>
| /* Given 'filename' (no extension) and 'ext', generate 'fname' */
| static void
| build_fname(char * fname, char const * filename, char const * ext)
| {
| if (fdopt.output_dir) {
| strcpy(fname, fdopt.output_dir);
| if (fname[strlen(fdopt.output_dir) - 1] != '/')
| strcat(fname, "/");
| strcat(fname, filename_only(filename));
| } else {
| size_t const buffer_size = 512;
| char buffer[buffer_size];
| char * cwd = getcwd(buffer, buffer_size);
| if (cwd) {
| strcpy(fname, cwd);
| strcat(fname, "/");
| strcat(fname, filename_only(filename));
| } else
| strcpy(fname, filename);
| }
| strcat(fname, ext);
| }
It almost seems to me that the else clause should just be
strcpy(fname, filename_only(filename));
and that you don't have to fiddle with getcwd at all?
| int main(int argc, char * argv[])
| {
| char fname[512];
Is this big enough?
(I have no idea what the actual value in xforms is, but PATH_MAX is a
nice value)
| fdopt.output_dir = 0;
| if (argc > 1) {
| fdopt.output_dir = (char *)malloc(strlen(argv[1]) + 1);
| strcpy(fdopt.output_dir, argv[1]);
| }
>
| build_fname(fname, "foo/bar/baz", ".h");
If my above method was used this would produce just baz.h. but the old
way would produce foo/bar/baz.h
--
Lgb
Re: fdesign -convert is buggy
Angus Leeming wrote:
> Lars Gullik Bjønnes wrote:
>> (and I am not saying that -dir should be removed, I am saying that
>> fdesign should output to pwd (unless -dir is specified of course.))
>
> Ok, understood. You want me to mess around with malloc just before
> we try and get XForms 1.1 out of the door. Bad man!
>
> I'll have a look.
This is the current code.
static char const *
filename_only(char const * filename)
{
char const * ptr = strrchr(filename, '/');
if (ptr)
return ptr+1;
return filename;
}
/* Given 'filename' (no extension) and 'ext', generate 'fname' */
static void
build_fname(char * fname, char const * filename, char const * ext)
{
if (fdopt.output_dir) {
strcpy(fname, fdopt.output_dir);
if (fname[strlen(fdopt.output_dir) - 1] != '/')
strcat(fname, "/");
strcat(fname, filename_only(filename));
} else
strcpy(fname, filename);
strcat(fname, ext);
}
I believe that we achieve what you're looking for with
-} else
- strcpy(fname, filename);
+} else {
+ size_t const buffer_size = 512;
+ char buffer[buffer_size];
+ char * cwd = getcwd(buffer, buffer_size);
+ if (cwd) {
+ strcpy(fname, cwd);
+ strcat(fname, "/");
+ strcat(fname, filename_only(filename));
+ } else
+ strcpy(fname, filename);
+}
Test code attached. I'm pretty sure that this is safe, but would value
second opinions...
--
Angus#include
#include
#include
typedef struct {
char * output_dir;
} FD_Opt;
FD_Opt fdopt;
static char const *
filename_only(char const * filename)
{
char const * ptr = strrchr(filename, '/');
if (ptr)
return ptr+1;
return filename;
}
/* Given 'filename' (no extension) and 'ext', generate 'fname' */
static void
build_fname(char * fname, char const * filename, char const * ext)
{
if (fdopt.output_dir) {
strcpy(fname, fdopt.output_dir);
if (fname[strlen(fdopt.output_dir) - 1] != '/')
strcat(fname, "/");
strcat(fname, filename_only(filename));
} else {
size_t const buffer_size = 512;
char buffer[buffer_size];
char * cwd = getcwd(buffer, buffer_size);
if (cwd) {
strcpy(fname, cwd);
strcat(fname, "/");
strcat(fname, filename_only(filename));
} else
strcpy(fname, filename);
}
strcat(fname, ext);
}
int main(int argc, char * argv[])
{
char fname[512];
fdopt.output_dir = 0;
if (argc > 1) {
fdopt.output_dir = (char *)malloc(strlen(argv[1]) + 1);
strcpy(fdopt.output_dir, argv[1]);
}
build_fname(fname, "foo/bar/baz", ".h");
printf("Output fname is \"%s\"\n", fname);
return 0;
}
Re: fdesign -convert is buggy
Lars Gullik Bjønnes wrote: > (and I am not saying that -dir should be removed, I am saying that > fdesign should output to pwd (unless -dir is specified of course.)) Ok, understood. You want me to mess around with malloc just before we try and get XForms 1.1 out of the door. Bad man! I'll have a look. -- Angus
Re: fdesign -convert is buggy
Angus Leeming <[EMAIL PROTECTED]> writes: | Jean-Marc Lasgouttes wrote: >> Lars> Jean-Marc Lasgouttes <[EMAIL PROTECTED]> writes: >> Lars> | This would break compatibility with older versions of >> fdesign. >> >> Lars> Would it really? >> >> Lars> Older versions of fdesign only works with the *.fd file in the >> Lars> same dir as where fdesign is run... >> >> Lars> So I see no compatibility problems. >> >> Good remark, it seems you are right. >> >> Angus, this probably means that we can/should get rid of the -dir >> argument. > | We would merely be swapping one semantic that works for another one | which is not yet proven. I don't see what it gains us. As I see it I | can do: Note that the old semantic is kept as well. Only if you do something that was not possible with old fdesign the semantics change. > | $ cd $HOME | $ fdesign -dir destdir -convert foo/bar.fd > | which seems more flexible to me than Lars' proposal. surely not. (and I am not saying that -dir should be removed, I am saying that fdesign should output to pwd (unless -dir is specified of course.)) -- Lgb
Re: fdesign -convert is buggy
Angus Leeming <[EMAIL PROTECTED]> writes: | Man page attached, but my conclusion is that uic is just different. | Nonetheless, the sematics that we have now have similar power to | those of uic. Except for the surprising behaviour that fdesign output files to the same dir there the .fd file is located. Sure, keep -dir, but make fdesign output to pwd not to the .fd files dir. -- Lgb
Re: fdesign -convert is buggy
Jean-Marc Lasgouttes wrote: >> "Angus" == Angus Leeming <[EMAIL PROTECTED]> writes: > > Angus> We would merely be swapping one semantic that works for > another Angus> one which is not yet proven. I don't see what it > gains us. As I Angus> see it I can do: > > Angus> $ cd $HOME $ fdesign -dir destdir -convert foo/bar.fd > > Angus> which seems more flexible to me than Lars' proposal. > > What does uic do? > > JMarc 1. It produces only one file per invocation, either .h or .C: Generate declaration: uic [options] Generate implementation: uic [options] -impl :name of the declaration file fdesign produces both .h and .c file in the same step. GENERAL OPTIONS -o file Write output to file rather than to stdout. Man page attached, but my conclusion is that uic is just different. Nonetheless, the sematics that we have now have similar power to those of uic. -- Angus uic.1.gz Description: GNU Zip compressed data
Re: fdesign -convert is buggy
> "Angus" == Angus Leeming <[EMAIL PROTECTED]> writes: Angus> We would merely be swapping one semantic that works for another Angus> one which is not yet proven. I don't see what it gains us. As I Angus> see it I can do: Angus> $ cd $HOME $ fdesign -dir destdir -convert foo/bar.fd Angus> which seems more flexible to me than Lars' proposal. What does uic do? JMarc
Re: fdesign -convert is buggy
Jean-Marc Lasgouttes wrote: > Lars> Jean-Marc Lasgouttes <[EMAIL PROTECTED]> writes: > Lars> | This would break compatibility with older versions of > fdesign. > > Lars> Would it really? > > Lars> Older versions of fdesign only works with the *.fd file in the > Lars> same dir as where fdesign is run... > > Lars> So I see no compatibility problems. > > Good remark, it seems you are right. > > Angus, this probably means that we can/should get rid of the -dir > argument. We would merely be swapping one semantic that works for another one which is not yet proven. I don't see what it gains us. As I see it I can do: $ cd $HOME $ fdesign -dir destdir -convert foo/bar.fd which seems more flexible to me than Lars' proposal. -- Angus
Re: fdesign -convert is buggy
Lars Gullik Bjønnes wrote:
> Angus Leeming <[EMAIL PROTECTED]> writes:
>
> | @@ -610,8 +600,14 @@ pre_connect(int ac, char *av[])
> |
> | for (s = i; s < ac; s++)
> | {
> | - if (load_forms(FALSE, av[s], 0) >= 0)
> | - save_forms(av[s]);
> | + if (load_forms(FALSE, av[s], 0) < 0) {
> | + fprintf(stderr, "Unable to load %s\n", av[s]);
> | + exit(1);
> | + }
> | + if (!save_forms(av[s])) {
> | + fprintf(stderr, "Unable to convert %s\n", av[s]);
> | + exit(1);
> | + }
>
> What if save_forms fails becuase it is unable to create the new
> files? But conversion proper when just ok?
The routine is misnamed. It does both (any necessary) saving and
conversion.
>
> | @@ -780,22 +771,8 @@ main(int ac, char *av[])
> | fd_buttonLabelSize = fl_cntl.buttonFontSize;
> | fl_cntl.buttonFontSize = 0;
> |
> | -
> | /* Initialize stuff */
> | init_classes();
> | -
> | -if (fdopt.conv_only)
> | -{
> | - if (ac == 1)
> | - fprintf(stderr, "%s: -convert requires arguments\n", av[0]);
> | -
> | - for (s = 1; s < ac; s++)
> | - {
> | - load_forms(FALSE, av[s], 0);
> | - save_forms(av[s]);
> | - }
> | - exit(0);
> | -}
>
> Isn't this removing functionality?
No. This code is never reached.
--
Angus
Re: fdesign -convert is buggy
> "Lars" == Lars Gullik Bjønnes <[EMAIL PROTECTED]> writes: Lars> Jean-Marc Lasgouttes <[EMAIL PROTECTED]> writes: Lars> | This would break compatibility with older versions of fdesign. Lars> Would it really? Lars> Older versions of fdesign only works with the *.fd file in the Lars> same dir as where fdesign is run... Lars> So I see no compatibility problems. Good remark, it seems you are right. Angus, this probably means that we can/should get rid of the -dir argument. JMarc
Re: fdesign -convert is buggy
Jean-Marc Lasgouttes <[EMAIL PROTECTED]> writes: >> "Lars" == Lars Gullik Bjønnes <[EMAIL PROTECTED]> writes: > | Lars> | Moreover, running (again from the builddir != src dir): >>> | Lars> | fdesign -dir . -convert | | Lars> /home/angus/lyx/devel/src/frontends/xforms/forms/form_external.fd > | Lars> is the -dir . needed? IMHO it shouldn't be. fdesign should just | Lars> output its files to PWD > | This would break compatibility with older versions of fdesign. Would it really? Older versions of fdesign only works with the *.fd file in the same dir as where fdesign is run... So I see no compatibility problems. -- Lgb
Re: fdesign -convert is buggy
> "Lars" == Lars Gullik Bjønnes <[EMAIL PROTECTED]> writes: Lars> | Moreover, running (again from the builddir != src dir): >> Lars> | fdesign -dir . -convert | Lars> /home/angus/lyx/devel/src/frontends/xforms/forms/form_external.fd Lars> is the -dir . needed? IMHO it shouldn't be. fdesign should just Lars> output its files to PWD This would break compatibility with older versions of fdesign. JMarc
Re: fdesign -convert is buggy
Angus Leeming <[EMAIL PROTECTED]> writes:
| @@ -610,8 +600,14 @@ pre_connect(int ac, char *av[])
|
| for (s = i; s < ac; s++)
| {
| - if (load_forms(FALSE, av[s], 0) >= 0)
| - save_forms(av[s]);
| + if (load_forms(FALSE, av[s], 0) < 0) {
| + fprintf(stderr, "Unable to load %s\n", av[s]);
| + exit(1);
| + }
| + if (!save_forms(av[s])) {
| + fprintf(stderr, "Unable to convert %s\n", av[s]);
| + exit(1);
| + }
What if save_forms fails becuase it is unable to create the new files?
But conversion proper when just ok?
| @@ -780,22 +771,8 @@ main(int ac, char *av[])
| fd_buttonLabelSize = fl_cntl.buttonFontSize;
| fl_cntl.buttonFontSize = 0;
|
| -
| /* Initialize stuff */
| init_classes();
| -
| -if (fdopt.conv_only)
| -{
| - if (ac == 1)
| - fprintf(stderr, "%s: -convert requires arguments\n", av[0]);
| -
| - for (s = 1; s < ac; s++)
| - {
| - load_forms(FALSE, av[s], 0);
| - save_forms(av[s]);
| - }
| - exit(0);
| -}
Isn't this removing functionality?
--
Lgb
Re: fdesign -convert is buggy
Angus Leeming <[EMAIL PROTECTED]> writes: | Do you have xforms cvs? no (and that is probably good since I now caught the problem myself, and didn't have to debug somebody elses problem) | If my testing doesn't throw up any problems, | I'm going to commit the attached patch. It gives you improved | diagnostics. nice > | The cvs version of fdesign will already enable you to run | $ fdesign -dir -convert form_aboutlyx.fd > | For example: > | Testing shows that running the cvs version of fdesign from the build | dir (!= src dir) so: | fdesign -convert | /home/angus/lyx/devel/src/frontends/xforms/forms/form_external.fd > | works, shoving form_external.[ch] in the same directory as the .fd | file. I don't agree with that behaviour. > | Moreover, running (again from the builddir != src dir): > | fdesign -dir . -convert | /home/angus/lyx/devel/src/frontends/xforms/forms/form_external.fd is the -dir . needed? IMHO it shouldn't be. fdesign should just output its files to PWD | shoves form_external.[ch] in the build dir. > | Ie, current cvs does all that you want of it. Good. But until we require xforms 1.1 we will need the work-around. -- Lgb
Re: fdesign -convert is buggy
Lars Gullik Bjnnes wrote:
> Angus Leeming <[EMAIL PROTECTED]> writes:
>
> | Lars Gullik Bjnnes wrote:
>>> It seems that fdesign returns 0 even if something went wrong.
>
> I forgot to say "when using -convert"
>
> | Wrong list...
>
> I know... but this is where I have the problem...
>
> | Both load_forms and save_forms return whether they were successful
> | or not, so this could easily be changed to:
>>
> | for (s = 1; s < ac; s++)
> | {
> | if (!load_forms(FALSE, av[s], 0))
> | exit(1);
> | if (!save_forms(av[s]))
> | exit(1);
> | }
>>
> | Thanks for the heads up.
>
> Sure.
>
> But some kind of error handling and a message to the user would be
> nice.
>
> And that it failes when the fd file is called as ./foo.fd is a bit
> strange...
Do you have xforms cvs? If my testing doesn't throw up any problems,
I'm going to commit the attached patch. It gives you improved
diagnostics.
The cvs version of fdesign will already enable you to run
$ fdesign -dir -convert form_aboutlyx.fd
For example:
Testing shows that running the cvs version of fdesign from the build
dir (!= src dir) so:
fdesign -convert
/home/angus/lyx/devel/src/frontends/xforms/forms/form_external.fd
works, shoving form_external.[ch] in the same directory as the .fd
file.
Moreover, running (again from the builddir != src dir):
fdesign -dir . -convert
/home/angus/lyx/devel/src/frontends/xforms/forms/form_external.fd
shoves form_external.[ch] in the build dir.
Ie, current cvs does all that you want of it.
--
AngusIndex: fdesign/fd_main.c
===
RCS file: /cvsroot/xforms/xforms/fdesign/fd_main.c,v
retrieving revision 1.10
diff -u -p -r1.10 fd_main.c
--- fdesign/fd_main.c 18 May 2004 13:57:11 - 1.10
+++ fdesign/fd_main.c 27 May 2004 12:09:10 -
@@ -86,12 +86,10 @@ FD_Opt fdopt;
Conv convertor[MAX_CONVERTOR + 1];
int fd_cntlborder;
-int help; /* if convert only */
int fd_bwidth;
int is_pasting;
int fd_trackgeometry = 1;
int fd_show_palette;
-int fd_pversion;
int fd_buttonLabelSize;
int fd_helpfontsize = 14;
int fd_align_fontsize = FL_TINY_SIZE;
@@ -587,15 +585,7 @@ pre_connect(int ac, char *av[])
if (fdopt.conv_only)
{
- fl_set_app_name(av[0], "Fdesign"); /* resource routine want this
-
-
-
-
-
-
-
- */
+ fl_set_app_name(av[0], "Fdesign"); /* resource routine wants this */
fl_init_fl_context();
create_the_forms();
init_classes();
@@ -610,8 +600,14 @@ pre_connect(int ac, char *av[])
for (s = i; s < ac; s++)
{
- if (load_forms(FALSE, av[s], 0) >= 0)
- save_forms(av[s]);
+ if (load_forms(FALSE, av[s], 0) < 0) {
+ fprintf(stderr, "Unable to load %s\n", av[s]);
+ exit(1);
+ }
+ if (!save_forms(av[s])) {
+ fprintf(stderr, "Unable to convert %s\n", av[s]);
+ exit(1);
+ }
}
exit(0);
}
@@ -721,7 +717,9 @@ main(int ac, char *av[])
initialize();
-/* For conversion and usage help, we don't need a connection */
+/* For conversion, version and usage help, we don't need a connection.
+ pre_connect will exit in such circumstances.
+*/
pre_connect(ac, av);
/* force fdesign to come up in default visual */
@@ -741,19 +739,12 @@ main(int ac, char *av[])
fl_get_app_resources(fdres, Nropt);
fl_add_signal_callback(SIGINT, interrupted, 0);
-if (fd_pversion)
- print_version(1);
-
-if (help)
- usage(av[0], 1);
-
if (av[1] && av[1][0] == '-')
{
fprintf(stderr, " Unknown option: %s\n", av[1]);
usage(av[0], 1);
}
-
fl_cntl.coordUnit = FL_COORD_PIXEL;
fdopt.unit = unit_val(fd_sunit);
fdopt.language = lang_val(fd_slanguage);
@@ -780,22 +771,8 @@ main(int ac, char *av[])
fd_buttonLabelSize = fl_cntl.buttonFontSize;
fl_cntl.buttonFontSize = 0;
-
/* Initialize stuff */
init_classes();
-
-if (fdopt.conv_only)
-{
- if (ac == 1)
- fprintf(stderr, "%s: -convert requires arguments\n", av[0]);
-
- for (s = 1; s < ac; s++)
- {
- load_forms(FALSE, av[s], 0);
- save_forms(av[s]);
- }
- exit(0);
-}
fl_set_counter_bounds(fd_align->snapobj, 0.0, 500.0);
fl_set_counter_step(fd_align->snapobj, 1.0, 5.0);
Re: fdesign -convert is buggy
Angus Leeming <[EMAIL PROTECTED]> writes:
| Lars Gullik Bjønnes wrote:
>> It seems that fdesign returns 0 even if something went wrong.
I forgot to say "when using -convert"
| Wrong list...
I know... but this is where I have the problem...
| Both load_forms and save_forms return whether they were successful or
| not, so this could easily be changed to:
>
| for (s = 1; s < ac; s++)
| {
| if (!load_forms(FALSE, av[s], 0))
| exit(1);
| if (!save_forms(av[s]))
| exit(1);
| }
>
| Thanks for the heads up.
Sure.
But some kind of error handling and a message to the user would be
nice.
And that it failes when the fd file is called as ./foo.fd is a bit
strange...
--
Lgb
Re: [patch] unneeded files
[EMAIL PROTECTED] (Lars Gullik Bjønnes) writes: | -dir destdir is not really needed, what is needed is the argument to | -convert to be any file with path. And fdesign to output to current working dir. -- Lgb
Re: [patch] unneeded files
Angus Leeming <[EMAIL PROTECTED]> writes:
| Lars Gullik Bjønnes wrote:
>> The distcheck target in newer automake is a bit stricter than in
>> previous versions. Among others it requires that when doing a VPATH
>> build, _nothing_ should change outside the build dir.
>>
>> The scripts in xforms/forms does not ahere to this rule.
>
| Hmmm. I guess that this is the problem then:
>
| (cd ${DIRNAME} && ${FDESIGN} -convert ${FDFILE}) ||
| {
| echo "\"${FDESIGN} -convert ${FDFILE}}\" failed. Please investigate."
| exit 1
| }
>
| IIRC, fdesign must be run in the directory containing the .fd file.
| That changes with XForms 1.1, allowing '-dir destdir' to be passed to
| fdesign so that it outputs all generated files in destdir. That's no
| help here of course because we still support XForms 1.0.
-dir destdir is not really needed, what is needed is the argument to
-convert to be any file with path.
| So I guess that you'll have to copy ${FDFILE} over to the build tree
| if build tree != src tree. You'll then be able to run fdesign in the
| build directory.
Not if I use ./*.fd ...
This should also be fixed in fdesign I guess.
--
Lgb
Re: fdesign -convert is buggy
Lars Gullik Bjønnes wrote: > And the arguemnt to convert _must_ be _just_ a file. Not even ./ is > allowed in front of the filename. I know. Crappy isn't it. -- Angus
Re: fdesign -convert is buggy
Lars Gullik Bjønnes wrote:
> It seems that fdesign returns 0 even if something went wrong.
>
> (writing the files f.ex.)
Wrong list...
In general, you're wrong. fdesign will exit early and return 1 in lots
of places:
if (!(fd_display = fl_initialize(&ac, av, 0, fd_cmdopt, Ncopt)))
exit(1);
Also, it calls 'usage' in lots of places. the '1' leads to exit(1).
if (help)
usage(av[0], 1);
But in this specific conversion case you're right, fdesign always
returns 0:
if (fdopt.conv_only)
{
if (ac == 1)
fprintf(stderr, "%s: -convert requires arguments\n", av[0]);
for (s = 1; s < ac; s++)
{
load_forms(FALSE, av[s], 0);
save_forms(av[s]);
}
exit(0);
}
Both load_forms and save_forms return whether they were successful or
not, so this could easily be changed to:
for (s = 1; s < ac; s++)
{
if (!load_forms(FALSE, av[s], 0))
exit(1);
if (!save_forms(av[s]))
exit(1);
}
Thanks for the heads up.
--
Angus
Re: fdesign -convert is buggy
[EMAIL PROTECTED] (Lars Gullik Bjønnes) writes: | It seems that fdesign returns 0 even if something went wrong. > | (writing the files f.ex.) And the arguemnt to convert _must_ be _just_ a file. Not even ./ is allowed in front of the filename. -- Lgb
Re: [patch] unneeded files
Lars Gullik Bjønnes wrote:
> The distcheck target in newer automake is a bit stricter than in
> previous versions. Among others it requires that when doing a VPATH
> build, _nothing_ should change outside the build dir.
>
> The scripts in xforms/forms does not ahere to this rule.
Hmmm. I guess that this is the problem then:
(cd ${DIRNAME} && ${FDESIGN} -convert ${FDFILE}) ||
{
echo "\"${FDESIGN} -convert ${FDFILE}}\" failed. Please investigate."
exit 1
}
IIRC, fdesign must be run in the directory containing the .fd file.
That changes with XForms 1.1, allowing '-dir destdir' to be passed to
fdesign so that it outputs all generated files in destdir. That's no
help here of course because we still support XForms 1.0.
So I guess that you'll have to copy ${FDFILE} over to the build tree
if build tree != src tree. You'll then be able to run fdesign in the
build directory.
--
Angus
fdesign -convert is buggy
It seems that fdesign returns 0 even if something went wrong. (writing the files f.ex.) -- Lgb
Re: [patch] unneeded files
Angus Leeming <[EMAIL PROTECTED]> writes: | Lars Gullik Bjønnes wrote: >> It would be nice if people could try this patch (the mkinstalldirs >> part really) and see if they have any problems. > | Lars, > | I've had no problems at all installing in /home/angus/testing | following 'make maintainer-clean'. This with a stock FC 1. Good. Now I need to fix up the "distcheck" target. The distcheck target in newer automake is a bit stricter than in previous versions. Among others it requires that when doing a VPATH build, _nothing_ should change outside the build dir. The scripts in xforms/forms does not ahere to this rule. -- Lgb
Re: [patch] unneeded files
Lars Gullik Bjønnes wrote: > It would be nice if people could try this patch (the mkinstalldirs > part really) and see if they have any problems. Lars, I've had no problems at all installing in /home/angus/testing following 'make maintainer-clean'. This with a stock FC 1. Angus -- Angus
Re: Why does 'make' trigger 'msgmerge'?
Lars Gullik Bjønnes wrote: > Angus Leeming <[EMAIL PROTECTED]> writes: > | It's never done this before. We've always needed 'make dist' or > | suchlike. > > Newer gettext work a bit differently. The merge is triggered when > lyx.pot is created and since we don't have it in CVS... Good oh. Glad to hear that all is going according to plan, for some meaning of plan. -- Angus
Re: Why does 'make' trigger 'msgmerge'?
Angus Leeming <[EMAIL PROTECTED]> writes: | It's never done this before. We've always needed 'make dist' or | suchlike. Newer gettext work a bit differently. The merge is triggered when lyx.pot is created and since we don't have it in CVS... > | This is this morning (so I've got Lars nt_defines.h patch) on a stock | FC1 after: > | $ make maintainer-clean | $ cd .. | $ cvs -q up | $ ./autogen.sh | $ mkdir build && cd build | $ ../configure --prefix=/home/angus/testing --enable-maintainer-mode | --with-version-suffix --with-included-gettext --with-frontend='xforms | qt gtk' | $ make > | Fortunately, subsequent 'make's don't trigger 'msgmerge. As designed by the gettext system. -- Lgb
Why does 'make' trigger 'msgmerge'?
It's never done this before. We've always needed 'make dist' or suchlike. This is this morning (so I've got Lars nt_defines.h patch) on a stock FC1 after: $ make maintainer-clean $ cd .. $ cvs -q up $ ./autogen.sh $ mkdir build && cd build $ ../configure --prefix=/home/angus/testing --enable-maintainer-mode --with-version-suffix --with-included-gettext --with-frontend='xforms qt gtk' $ make Fortunately, subsequent 'make's don't trigger 'msgmerge. Still building. Will report back later whether I installed successfully in /home/angus/testing. -- Angus
[patch] unneeded files
It would be nice if people could try this patch (the mkinstalldirs
part really) and see if they have any problems.
My testing so far shows no problems.
(OTOH I know of a problem in POTFILES.in generation...
src/support/path_defines.C is not always there... and should thus be
excluded from POTFILES.in but isn't)
? Config
? counter-2.diff
? counter-3.diff
? counter.diff
? deptherror.diff
? deptherror.lyx
? dispatch-2.diff
? dispatch-3.diff
? dispatch-4.diff
? end.lyx
? kystskipper-a-1.lyx
? morectrs.lyx
? refenum.lyx
? unneeded.diff
Index: mkinstalldirs
===
RCS file: mkinstalldirs
diff -N mkinstalldirs
--- mkinstalldirs 21 Nov 2000 16:35:36 - 1.1
+++ /dev/null 1 Jan 1970 00:00:00 -
@@ -1,32 +0,0 @@
-#! /bin/sh
-# mkinstalldirs --- make directory hierarchy
-# Author: Noah Friedman <[EMAIL PROTECTED]>
-# Created: 1993-05-16
-# Public domain
-
-errstatus=0
-
-for file
-do
- set fnord `echo ":$file" | sed -ne 's/^:\//#/;s/^://;s/\// /g;s/^#/\//;p'`
- shift
-
- pathcomp=
- for d in ${1+"$@"} ; do
- pathcomp="$pathcomp$d"
- case "$pathcomp" in
- -* ) pathcomp=./$pathcomp ;;
- esac
-
- if test ! -d "$pathcomp"; then
-echo "mkdir $pathcomp" 1>&2
-mkdir "$pathcomp" || errstatus=$?
- fi
-
- pathcomp="$pathcomp/"
- done
-done
-
-exit $errstatus
-
-# mkinstalldirs ends here
Index: config/depcomp.diff
===
RCS file: config/depcomp.diff
diff -N config/depcomp.diff
--- config/depcomp.diff 22 Mar 2002 12:32:11 - 1.1
+++ /dev/null 1 Jan 1970 00:00:00 -
@@ -1,39 +0,0 @@
-Index: depcomp
-===
-RCS file: /usr/local/lyx/cvsroot/lyx-devel/config/depcomp,v
-retrieving revision 1.1
-diff -u -p -r1.1 depcomp
depcomp 17 Oct 2001 15:38:58 - 1.1
-+++ depcomp 22 Mar 2002 12:18:23 -
-@@ -198,14 +198,28 @@ aix)
- ;;
-
- tru64)
-- # The Tru64 AIX compiler uses -MD to generate dependencies as a side
-+ # The Tru64 DEC compiler uses -MD to generate dependencies as a side
-# effect. `cc -MD -o foo.o ...' puts the dependencies into `foo.o.d'.
-# At least on Alpha/Redhat 6.1, Compaq CCC V6.2-504 seems to put
-# dependencies in `foo.d' instead, so we check for that too.
-# Subdirectories are respected.
-
-- tmpdepfile1="$object.d"
-- tmpdepfile2=`echo "$object" | sed -e 's/.o$/.d/'`
-+ # This is incorrect as it assumes that $object has a .o extension when
-+ # it could well have a .lo one.
-+ #tmpdepfile1="$object.d"
-+ # This is just bad sed. The /.o$/ should be /\.o/, nonwithstanding the
-+ # flawed logic, described above.
-+ #tmpdepfile2=`echo "$object" | sed -e 's/.o$/.d/'`
-+
-+ # The correct way:
-+ # Strip the (.o, or .lo) extension from $object
-+ # (Same piece of sed magic as used for the AIX compiler above.)
-+ stripped=`echo "$object" | sed -e 's,^.*/,,' -e 's/\(.*\)\..*$/\1/'`
-+ # Tru64 compiler dependency file.
-+ tmpdepfile1="$stripped.o.d"
-+ # Compaq CCC V6.2 dependency file.
-+ tmpdepfile2="$stripped.d"
-+
-if test "$libtool" = yes; then
- "$@" -Wc,-MD
-else
Index: config/gettext.m4.patch
===
RCS file: config/gettext.m4.patch
diff -N config/gettext.m4.patch
--- config/gettext.m4.patch 8 Jan 2000 21:06:59 - 1.1
+++ /dev/null 1 Jan 1970 00:00:00 -
@@ -1,16 +0,0 @@
/usr/share/aclocal/gettext.m4 Thu Sep 23 05:05:24 1999
-+++ gettext.m4 Sat Jan 8 04:12:49 2000
-@@ -227,7 +227,12 @@
- else
-AC_MSG_CHECKING(for catalogs to be installed)
-NEW_LINGUAS=
-- for lang in ${LINGUAS=$ALL_LINGUAS}; do
-+dnl == Fix is here! ==
-+dnl Some sh do not like substitution in bounds of for loops
-+dnlfor lang in ${LINGUAS=$ALL_LINGUAS}; do
-+tmplinguas=${LINGUAS=$ALL_LINGUAS}
-+for lang in ${tmplinguas}; do
-+dnl end of fix
- case "$ALL_LINGUAS" in
- *$lang*) NEW_LINGUAS="$NEW_LINGUAS $lang" ;;
- esac
Index: po/.cvsignore
===
RCS file: /usr/local/lyx/cvsroot/lyx-devel/po/.cvsignore,v
retrieving revision 1.10
diff -u -p -r1.10 .cvsignore
--- po/.cvsignore 26 May 2004 16:53:46 - 1.10
+++ po/.cvsignore 27 May 2004 08:24:59 -
@@ -14,3 +14,4 @@ xforms_l10n.pot
default_ui_l10n.pot
languages_l10n.pot
ui_l10n.pot
+remove-potcdate.sed
Index: po/POTFILES.in
===
RCS file: /usr/local/lyx/cvsroot/lyx-devel/po/POTFILES.in,v
retrieving revision 1.391
diff -u -p -r1.391 POTFILES.in
--- po/POTFILES.in 26 May 2004 20:26:09 - 1.391
+++ po/POTFILES.in 27 May 2004 08:24:59 -
@@ -188,6 +188,7 @@ src/output_plaintext.C
src/paragraph.C
src/rowpainter.C
src/support/globbing.C
+src/support/path_defines.C
s
Re: The real solution to the Cygwin compile problems
"Kayvan A. Sylvan" <[EMAIL PROTECTED]> writes: | I got tired of wrestling with the config mechanism. > | Please apply this (support/nt_defines.h is not used) and then | "cvs remove" the support/nt_defines.h file. I'll handle this one. -- Lgb
The real solution to the Cygwin compile problems
I got tired of wrestling with the config mechanism. Please apply this (support/nt_defines.h is not used) and then "cvs remove" the support/nt_defines.h file. Thanks. -- Kayvan A. Sylvan Sylvan Associates, Inc. http://sylvan.com/~kayvan Index: ChangeLog === RCS file: /cvs/lyx/lyx-devel/ChangeLog,v retrieving revision 1.964 diff -u -r1.964 ChangeLog --- ChangeLog 2004/05/26 23:03:07 1.964 +++ ChangeLog 2004/05/27 07:07:02 @@ -1,3 +1,9 @@ +2004-05-27 Kayvan Sylvan <[EMAIL PROTECTED]> + + * configure.ac: Remove reference to support/nt_defines.h + + * config/acconfig.h: Remove reference to support/nt_defines.h + 2004-05-26 Lars Gullik Bjonnes <[EMAIL PROTECTED]> * autogen.sh (ACLOCAL): make all aclocal find m4 Index: configure.ac === RCS file: /cvs/lyx/lyx-devel/configure.ac,v retrieving revision 1.19 diff -u -r1.19 configure.ac --- configure.ac2004/05/26 16:53:44 1.19 +++ configure.ac2004/05/27 07:07:02 @@ -361,10 +361,6 @@ #include "support/os2_defines.h" #endif -#if defined(__CYGWIN__) || defined(__CYGWIN32__) -#include "support/nt_defines.h" -#endif - #if defined(HAVE_OSTREAM) && defined(HAVE_LOCALE) && defined(HAVE_SSTREAM) #define USE_BOOST_FORMAT 1 #else Index: config/acconfig.h === RCS file: /cvs/lyx/lyx-devel/config/acconfig.h,v retrieving revision 1.2 diff -u -r1.2 acconfig.h --- config/acconfig.h 2002/11/24 15:20:31 1.2 +++ config/acconfig.h 2004/05/27 07:07:02 @@ -78,10 +78,6 @@ #include "support/os2_defines.h" #endif -#if defined(__CYGWIN__) || defined( __CYGWIN32__) -#include "support/nt_defines.h" -#endif - /* Some support for the boost library. */ /* * #ifndef HAVE_LIMITS Index: src/support/ChangeLog === RCS file: /cvs/lyx/lyx-devel/src/support/ChangeLog,v retrieving revision 1.253 diff -u -r1.253 ChangeLog --- src/support/ChangeLog 2004/05/20 09:35:30 1.253 +++ src/support/ChangeLog 2004/05/27 07:07:02 @@ -1,3 +1,7 @@ +2004-05-27 Kayvan Sylvan <[EMAIL PROTECTED]> + + * Makefile.am: remove reference to nt_defines.h + 2004-05-20 Angus Leeming <[EMAIL PROTECTED]> * debugstream.h: wrap the definition of debug_trait::DEBUG with Index: src/support/Makefile.am === RCS file: /cvs/lyx/lyx-devel/src/support/Makefile.am,v retrieving revision 1.73 diff -u -r1.73 Makefile.am --- src/support/Makefile.am 2004/01/06 19:32:05 1.73 +++ src/support/Makefile.am 2004/05/27 07:07:02 @@ -51,7 +51,6 @@ lyxtime.h \ lyxsum.C \ mkdir.C \ - nt_defines.h \ os.C \ os.h \ os2_defines.h \
Re: Questions on Instant Preview in 1.3.5
On Wednesday 26 May 2004 16:47, Jean-Marc Lassitude's wrote: > > Angus> The python script does all the manipulation using pure python > Angus> code, so it should "just work" if you have python installed. > > Do you think it works with ancient python like 1.5? From what I remember from the code the answer is yes. > JMarc PS: I don't know when this message will hit the list, but here it goes nevertheless. -- José Abílio LyX and docbook, a perfect match. :-)
