Re: Error with make doc

[Werner LEMBERG]

Folks,


Carl wrote to me privately:

> OK, so now I have a failure:
>
> I did
>
> ```
> rm -rf build
> mkdir build
> cd build
> ../autogen.sh --currdir
> ../configure &> configure.log
> make -j10 CPU_COUNT=10 all
> make doc LANGS="en" VERBOSE=1 CPU_COUNT=10 -j10 &> make.doc.log
> ```
> And the make doc failed.

[This is based on the branch 'dev/wl/build-fixes'.]

The log file `snippet-names-c5b165525229969dd268a5b19f9f64fd.log`
contains

```
GNU LilyPond 2.25.14 (running Guile 3.0)

Forking into jobs:  (39647 39646 39645 39644 39643 39642 39641 39640 39639 
39638)
job 9 terminated with signal: 6
job 8 terminated with signal: 6
job 7 terminated with signal: 6
job 6 terminated with signal: 6
job 5 terminated with signal: 6
job 4 terminated with signal: 6
job 3 terminated with signal: 6
job 2 terminated with signal: 6
job 1 terminated with signal: 6
job 0 terminated with signal: 6
fatal error: Children (9 8 7 6 5 4 3 2 1 0) exited with errors.
```

AFAIK, signal 6 is 'SIGABRT'.

Note that the files `lilypond-multi-run-*.log` all contain

  thread_suspend failed

which seems to be a hint that we have a problem with thread support on
macOS (Carl runs on macOS Sonoma which identifies itself as Darwin
23.2 on arm64e).

Any idea how to tackle this?  BTW, `make doc` without `CPU_COUNT=10`
succeeds on his computer.


Werner

Re: Error with make doc

[Werner LEMBERG]

> I can confirm that I get the proper HTML pages, compiling with -j10
> on both make bytecode and make doc.

Great!

> It was delightful to see the proper lilypond website on my local
> file system!  Thank you so much for fixing this!

Glad that I could help.  I will now polish the MR as requested by
Jonas.

>> The last step – I hope – will be to find out why `CPU_COUNT` fails...
>>
> 
> I will see if I can make CPU_COUNT fail.  If not, I guess the build
> system is good on MacOS!

The actual question is whether it works at all, i.e., whether setting
`CPU_COUNT=10` makes your system really compile multiple `.ly` files
in parallel.  Given that I've seen

  thread_suspend failed

in your log files I suspect that LilyPOnd simply switches to serial
processing instead.  If you don't get a timing difference (which means
that my assumption is correct) I will prepare something more specific
for further tests.


Werner

Re: Error with make doc

[Carl Sorensen]

On Mon, Feb 12, 2024 at 9:22 AM Werner LEMBERG  wrote:

>
>   v2.25.13-8-gbe52228a70
>
> Please test again!  Theoretically, you should now get proper HTML
> pages.
>

I can confirm that I get the proper HTML pages, compiling with -j10 on both
make bytecode and make doc.

It was delightful to see the proper lilypond website on my local file
system!  Thank you so much for fixing this!

>
> The last step – I hope – will be to find out why `CPU_COUNT` fails...
>

I will see if I can make CPU_COUNT fail.  If not, I guess the build system
is good on MacOS!

Thanks so much for your work on this!

Carl

Re: Error with make doc

[Carl Sorensen]

On Sat, Feb 10, 2024 at 7:20 AM Werner LEMBERG  wrote:

>
> >> systems like macOS that neither have or use '/bin/bash'.
> >
> > Apple changed the default shell for new users to zsh about 5 years
> > ago, but bash is still there.  I don't oppose making the build
> > scripts more portable, but let's not misunderstand the situation.
>
> OK, thanks – let's see whether the new code makes a build succeed on
> Carl's computer; otherwise I have to continue my investigations.
>
> Note that on OpenBSD, `bash` resides in `/usr/local/bin` if you
> install it via `pkg_add`.
>
>
On my system, default bash is at

/opt/local/bin/bash

But I do have

/bin/bash

Carl

Re: Error with make doc

[Werner LEMBERG]

>> systems like macOS that neither have or use '/bin/bash'.
> 
> Apple changed the default shell for new users to zsh about 5 years
> ago, but bash is still there.  I don't oppose making the build
> scripts more portable, but let's not misunderstand the situation.

OK, thanks – let's see whether the new code makes a build succeed on
Carl's computer; otherwise I have to continue my investigations.

Note that on OpenBSD, `bash` resides in `/usr/local/bin` if you
install it via `pkg_add`.


Werner

Re: Error with make doc

[Dan Eble]

On 2024-02-10 07:52, Werner LEMBERG wrote:


systems like macOS that neither have or use '/bin/bash'.


Apple changed the default shell for new users to zsh about 5 years ago, 
but bash is still there.  I don't oppose making the build scripts more 
portable, but let's not misunderstand the situation.

--
Dan

Re: Error with make doc

[Werner LEMBERG]

>> If I click on the Notation link from that page, I get to:
>>
>> out-www/offline-root/Documentation/web/notation.html
>>
>> which has no css, but has lots of links.
>
> This shouldn't happen, and it means that your documentation build is
> still not correct.

Continuing with the analysis of Carl's build tarball it seems that the
script `build-doc-tree.sh` doesn't work as expected.  I've now
converted it to a normal shell script (i.e., it no longer uses syntax
specific to bash), which should hopefully increase its portability to
systems like macOS that neither have or use '/bin/bash'.

Carl, please do another run of the following commands based on the
updated 'dev/wl/build-fixes' branch, then create and upload a tarball
of your `build` directory as before and send me a link to it.

```
./configure
make bytecode -j10 LANGS="en"
make doc LANGS="en"
```

Regarding the failing `make doc -j10`:

> > "/Users/carl/Development/lilypond/build/out/lybook-db/
> > snippet-names-c5b165525229969dd268a5b19f9f64fd.ly"' returned
> > non-zero exit status 1.
>
> Please try to find out which of the snippet files mentioned in the
> above file failed.  `*.log` files should be in directory
> `out/lybook-db`, too, and the error message should be near the end
> of one of the log files.
>
> I would need both the contents of the failing snippet file and the
> corresponding error message.

I still need this, BTW.


Werner

Re: Error with make doc

[Werner LEMBERG]

> If I click on the Notation link from that page, I get to:
> 
> out-www/offline-root/Documentation/web/notation.html
> 
> which has no css, but has lots of links.

This shouldn't happen, and it means that your documentation build is
still not correct.

> I'm happy to send a tar file of my offline-root directory and
> subdirectories if that would be helpful.

Yes, something like that would help.  However, I would like to see
your *complete* (successful) build of the `bytecode` and `doc` make
targets, with `LANGS="en"` to reduce space.  If you don't mind, please
issue the following command one level above your build directory:

  gtar cJvf build.tar.xz build

where `gtar` is GNU tar (from the 'gnutar' MacPorts package), also
assuming that your build directory is called `build`.  Upload the
created file `build.tar.xz` to a filesharing hoster of your choice and
send me the link privately so that I can download the file for further
inspection.  For me, this archive has about 206MByte.

[The `J` command makes GNU tar use the `xz` program for compression,
 which is much more efficient than the default `gz` (at the cost of
 compression speed), and which you probably have to install from
 MacPorts, too.]


Werner

Re: Error with make doc

[Werner LEMBERG]

> "/Users/carl/Development/lilypond/build/out/lybook-db/
> snippet-names-c5b165525229969dd268a5b19f9f64fd.ly"' returned
> non-zero exit status 1.

Please try to find out which of the snippet files mentioned in the
above file failed.  `*.log` files should be in directory
`out/lybook-db`, too, and the error message should be near the end of
one of the log files.

I would need both the contents of the failing snippet file and the
corresponding error message.


Werner

Re: Error with make doc

[Carl Sorensen]

On Sun, Feb 4, 2024 at 7:21 AM Werner LEMBERG  wrote:

>
>
> Please test
>
>   https://gitlab.com/lilypond/lilypond/-/merge_requests/2254
>
> It should fix all the build problems you've reported.
>

This merge request allowed me to build the documentation using MacOS cp and
find.

I still got errors if I tried make doc with -j12.  But without -j12 it
succeeded.

There are some minor issues with either the way the documentation is built
or the way I am using it.

IIRC, in the past, I could use Documentation/out-www/index.html as a
beginning place for documentation, and I could get to all the manuals.

Now, when I follow this link, I get into a documentation tree that has
links for all the manuals, but no css nor images. It forwards me to

out-www/offline-root/Documentation/web/index.html

If I click on the Notation link from that page, I get to:

out-www/offline-root/Documentation/web/notation.html

which has no css, but has lots of links.

Notation (split html) takes me to

out-www/offline-root/Documentation/notation/index.html

which also has no css, nor body text, nor images.

On the notation.html page, if I click on the Notation (big html) page, I
get to

out-www/offline-root/Documentation/notation-big-page.html

Which has appropriate css.

If instead of the above, I go to  the en tree version of Notation, I get

Documentation/out-www/en/notation/index.html

which has css, but no images.

When I click on the "Back to documentation index" from that page, I get to

Documentation/out-www/Documentation/web/manuals.html

WHich does not exist.

So it seems the one-big-page html docs have css and images, but not the
single pages.

I'm happy to send a tar file of my offline-root directory and
subdirectories if that would be helpful.

Thanks,

Carl

Re: Error with make doc

[Carl Sorensen]

On Sat, Feb 3, 2024 at 12:01 AM Werner LEMBERG  wrote:

>
> [Carl sent me the file listing off-list, thanks!  I will analyze it in
> due course.]
>
> > [...] and make test with a single job (to avoid race conditions that
> > have given me trouble in the past):
>
> Interesting.  When is the 'past'?  This shouldn't happen today, and it
> would be great if you could try a `-jXX' make option in the future so
> that we can see if you really get a failure.
>

It was current.  And it remains current with the latest merge request.


>
> > now on to your requested steps:
>
> Ah, my steps were not intended to be used additionally but instead :-)
>

Ah, so I tried that.
```
../configure
make bytecode -j12
```
Ran without error.

```

carl@carls-mbp-2 build % make doc -j12 CPU_COUNT=12 LANGS="en"

Making input/regression/out-www/collated-files.list < 1888 files

Making input/regression/out-www/collated-files.tely

Making input/regression/out-www/collated-files.texi < tely

Command '/Users/carl/Development/lilypond/build/out/bin/lilypond -I
/Users/carl/Development/lilypond/input/regression
-dseparate-page-formats=ps,pdf
-djob-count=12 -dinclude-eps-fonts -dgs-load-fonts --header=doctitle
--header=doctitleen --header=texidoc --header=texidocen
-dcheck-internal-types -danti-alias-factor=2
-dfont-ps-resdir=/Users/carl/Development/lilypond/build/out-fonts -O TeX-GS
-I  "/Users/carl/Development/lilypond/input/regression"  -I
"/Users/carl/Development/lilypond/build/input/regression"
-daux-files  -dread-file-list -dno-strip-output-dir
-dtall-page-formats=eps,png
"/Users/carl/Development/lilypond/build/out/lybook-db/
snippet-names-c5b165525229969dd268a5b19f9f64fd.ly"' returned non-zero exit
status 1.

make[3]: *** [out-www/collated-files.texi] Error 1

make[2]: *** [WWW] Error 2

make[1]: *** [WWW] Error 2
make: *** [doc] Error 2
```
I'm currently in the midst of a

```
make doc LANGS="en"
```
run.  I think it will complete without error, but it's slow...

If you tell me what I can run to provide information about the j12 error on
make doc, I'm happy to do that and send you whatever you ask for.

I'll send another email shortly with  the results of the make doc run with
only a single job.

Thanks,

Carl

Re: Error with make doc

[Werner LEMBERG]

>> > I omitted the '-l' option in the cp calls in the patch.
>> >
>> > 'make doc' succeeded without any errors, so I thought I had success.  But
>> > it turns out that
>> >
>> > a) I appear not to have the .css files copied (at least the output
>> >showed no css -- just basic white screen), and
>> > b) I appear not to have the image files copied (there are no images
>> >that show up in any of the pages accessible from
>> >www-out/offline-root).

As it turned out, this is another incompatibility: `cp -a` doesn't
work on macOS.

Please test

  https://gitlab.com/lilypond/lilypond/-/merge_requests/2254

It should fix all the build problems you've reported.


Werner

Re: Error with make doc

[Werner LEMBERG]

[Carl sent me the file listing off-list, thanks!  I will analyze it in
due course.]

> [...] and make test with a single job (to avoid race conditions that
> have given me trouble in the past):

Interesting.  When is the 'past'?  This shouldn't happen today, and it
would be great if you could try a `-jXX' make option in the future so
that we can see if you really get a failure.

> now on to your requested steps:

Ah, my steps were not intended to be used additionally but instead :-)

> Everything completed without error.

This is good.

Regarding the `cp -l` issue: Irrespective of the faulty implementation
on macOS it turns out that option `-l` isn't portable at all, as I was
told by the autoconf people.  I'm going to replace it with something
else that also works across file system boundaries (which even a
working `cp -l` does not).


Werner

Re: Error with make doc

[Carl Sorensen]

On Fri, Feb 2, 2024 at 4:34 AM Werner LEMBERG  wrote:

> >> Please read this and check whether you are affected by this
> >> problem.  In case you can confirm the issue please try whether it
> >> works if you omit the `-l` option in all `cp` calls in my patch.
> >
> > I checked -- the results of the problem were different for me than
> > for the bug report.  I got the error message, but the hard link was
> > never copied (there were not two files with the same inode -- only
> > the original file.
>
> OK, so there is indeed a problem with `cp -l` on your macOS.  It's
> still hard to believe.
>

Yes, and apparently it's a new bug (older versions of the OS didn't have
it, IIUC).


>
> BTW, which macOS version are you using?
>

 Sonoma 14.2.1 (23C71)

>
> > I omitted the '-l' option in the cp calls in the patch.
> >
> > 'make doc' succeeded without any errors, so I thought I had success.  But
> > it turns out that
> >
> > a) I appear not to have the .css files copied (at least the output
> >showed no css -- just basic white screen), and
> > b) I appear not to have the image files copied (there are no images
> >that show up in any of the pages accessible from
> >www-out/offline-root).
>
> It seems that my patch doesn't work as expected, or that I have missed
> something.  As usual, please start from scratch and run the standard
> incantation (or your adapted variation of it)
>



>
> ```
> ./configure
> make bytecode -j12
> make doc -j12 CPU_COUNT=12 LANGS="en"
> ```
>
> (please tell me what git commit you are using so that I can try
> exactly the same setup).  After finishing, please call
>
> ```
> find . | gzip > find.log.gz
> ```
>
> in your build directory and send me the file *privately* for further
> inspection (it's about 750kByte compressed – listing approx. 16
> files) and of no real interest to other readers of this list).
>
>
I will do this today.


> > Does it make sense to try again with the xargs part of the patch,
> > but using the GNU cp and find?
>
> For you personally, it certainly makes sense so that you get a
> successful build.  For the LilyPond project, however, I think it's
> best to use native OS tools so that the number of additional
> requisites stays as small as possible.
>

I tried with the xargs part  of the patch, with the GNU cp and find.  It
executed sucessfully (no errors).  But it also had no css on the resulting
index.html.

I viewed the source of index.html.  It referenced the css from
css/lilypond-website.css, but there was no 'css' directory in
build/out-www/offline-root/.   I tried copying the css directory from
lilypond/Documentation to build/out-www/offline-root, but still didn't get
a properly styled webpage.

Thanks,

Carl

Re: Error with make doc

[Werner LEMBERG]

>> Please read this and check whether you are affected by this
>> problem.  In case you can confirm the issue please try whether it
>> works if you omit the `-l` option in all `cp` calls in my patch.
>
> I checked -- the results of the problem were different for me than
> for the bug report.  I got the error message, but the hard link was
> never copied (there were not two files with the same inode -- only
> the original file.

OK, so there is indeed a problem with `cp -l` on your macOS.  It's
still hard to believe.

BTW, which macOS version are you using?

> I omitted the '-l' option in the cp calls in the patch.
>
> 'make doc' succeeded without any errors, so I thought I had success.  But
> it turns out that
>
> a) I appear not to have the .css files copied (at least the output
>showed no css -- just basic white screen), and
> b) I appear not to have the image files copied (there are no images
>that show up in any of the pages accessible from
>www-out/offline-root).

It seems that my patch doesn't work as expected, or that I have missed
something.  As usual, please start from scratch and run the standard
incantation (or your adapted variation of it)

```
./configure
make bytecode -j12
make doc -j12 CPU_COUNT=12 LANGS="en"
```

(please tell me what git commit you are using so that I can try
exactly the same setup).  After finishing, please call

```
find . | gzip > find.log.gz
```

in your build directory and send me the file *privately* for further
inspection (it's about 750kByte compressed – listing approx. 16
files) and of no real interest to other readers of this list).

> Does it make sense to try again with the xargs part of the patch,
> but using the GNU cp and find?

For you personally, it certainly makes sense so that you get a
successful build.  For the LilyPond project, however, I think it's
best to use native OS tools so that the number of additional
requisites stays as small as possible.


Werner

Re: Error with make doc

[Carl Sorensen]

On Thu, Feb 1, 2024 at 10:47 AM Werner LEMBERG  wrote:

>
>
> Doing a bit of an internet search it seems this is a bug in the `cp`
> implementation on macOS (as unbelievable it may be, given that `cp` is
> such a fundamental command that exists since decades):
>
>
> https://apple.stackexchange.com/questions/464706/why-do-i-get-bad-file-descriptor-when-copying-using-hardlink-flag
>
> Please read this and check whether you are affected by this problem.
> In case you can confirm the issue please try whether it works if you
> omit the `-l` option in all `cp` calls in my patch.
>

I checked -- the results of the problem were different for me than for the
bug report.  I got the error message, but the hard link was never copied
(there were not two files with the same inode -- only the original file.

I omitted the '-l' option in the cp calls in the patch.

'make doc' succeeded without any errors, so I thought I had success.  But
it turns out that

a) I appear not to have the .css files copied (at least the output showed
no css -- just basic white screen), and
b) I appear not to have the image files copied (there are no images that
show up in any of the pages accessible from www-out/offline-root).

Does it make sense to try again with the xargs part of the patch, but using
the GNU cp and find?

Carl

Re: Error with make doc

[Werner LEMBERG]

> 'make doc' failed.
> 
> It appears that under some conditions, we end up with a path
> including '//'.

Multiple slashes in succession are completely harmless; they are
simply equivalent to `/`.

Doing a bit of an internet search it seems this is a bug in the `cp`
implementation on macOS (as unbelievable it may be, given that `cp` is
such a fundamental command that exists since decades):

  
https://apple.stackexchange.com/questions/464706/why-do-i-get-bad-file-descriptor-when-copying-using-hardlink-flag

Please read this and check whether you are affected by this problem.
In case you can confirm the issue please try whether it works if you
omit the `-l` option in all `cp` calls in my patch.


Werner

Re: Error with make doc

[Carl Sorensen]

On Wed, Jan 31, 2024 at 10:53 PM Werner LEMBERG  wrote:

>
>
> Thanks a lot, very helpful.  Please apply the attached patch to the
> top-level `GNUmakefile.in` file, then retry from scratch (including
> the `configure` step); it should catch all of the three problems
> you've encountered.
>

It seems to work, but not perfectly.  'make all' and 'make test' work.
'make doc' failed.

It appears that under some conditions, we end up with a path including '//'.

Log file attached.

Thanks,

Carl


make.doc.log.gz
Description: GNU Zip compressed data

Re: Error with make doc

[Werner LEMBERG]

> I've attached three compressed log files.  [...]

Thanks a lot, very helpful.  Please apply the attached patch to the
top-level `GNUmakefile.in` file, then retry from scratch (including
the `configure` step); it should catch all of the three problems
you've encountered.

The `-maxdepth` and `-mindepth` options of `find` are not completely
portable according to the 'autoconf' info pages; however, we are using
them in `aclocal.m4` since ages and nobody complained, so...


Werner
diff --git a/GNUmakefile.in b/GNUmakefile.in
index 6c0e1f287d..b28130e779 100644
--- a/GNUmakefile.in
+++ b/GNUmakefile.in
@@ -153,7 +153,7 @@ $(outdir)/%-root/index.html: doc-build
 	mkdir -p $(dir $@)/Documentation/
 	for d in . musicxml abc2ly lilypond-book ; do \
 		mkdir -p $(dir $@)/input/regression/$${d}; \
-		cp --link -a input/regression/$${d}/out-www/*  $(dir $@)/input/regression/$${d}/ ; \
+		cp -l -a input/regression/$${d}/out-www/*  $(dir $@)/input/regression/$${d}/ ; \
 	done
 	find $(dir $@)/input/regression/ \
 		-type f \
@@ -164,18 +164,23 @@ $(outdir)/%-root/index.html: doc-build
 		-not -name '*.mxl' \
 		-not -name 'collated-files.pdf' \
 		-delete
-	cp --link -a Documentation/out-www/webdoc/* $(dir $@)/Documentation/
-	cp --link $(TOPDOC_TXT_FILES) $(dir $@)
+	cp -l -a Documentation/out-www/webdoc/* $(dir $@)/Documentation/
+	cp -l $(TOPDOC_TXT_FILES) $(dir $@)
 	# No link because the files come from the source directory.
 	cp $(addprefix $(src-dir)/,$(README_FILES)) $(dir $@)
 	$(PYTHON) $(buildscript-dir)/www_post.py $(PACKAGE_NAME) $(TOPLEVEL_VERSION) $(dir $@) "$*"
 	find $(outdir)/$*-root -type l | xargs rm -f
-	cd $(outdir)/$*-root && find -name \*.html -execdir $(buildscript-dir)/fix-docsize.sh {} +
-	$(PERL) -i -pe 's#(href)="\.\./Documentation/#\1="#g' $(dir $@)/Documentation/*html
-	$(PERL) -i -pe 's#(href)="\.\./\.\./Documentation/#\1="../#g' $(dir $@)/Documentation/*/*html
-	# Fixup links to CSS files.
-	$(PERL) -i -pe 's#(href)="css/#\1="../../Documentation/css/#g' $(dir $@)/input/regression/*html
-	$(PERL) -i -pe 's#(href)="css/#\1="../../../Documentation/css/#g' $(dir $@)/input/regression/*/*html
+	cd $(outdir)/$*-root \
+	  && find . -name "*.html" -execdir $(buildscript-dir)/fix-docsize.sh {} +
+	find $(dir $@)/Documentation -maxdepth 1 -name "*html" \
+	  | xargs $(PERL) -i -pe 's#(href)="\.\./Documentation/#\1="#g'
+	find $(dir $@)/Documentation -mindepth 2 -maxdepth 2 -name "*html" \
+	  | xargs $(PERL) -i -pe 's#(href)="\.\./\.\./Documentation/#\1="../#g'
+	# Fix up links to CSS files.
+	find $(dir $@)/input/regression -maxdepth 1 -name "*html" \
+	  | xargs $(PERL) -i -pe 's#(href)="css/#\1="../../Documentation/css/#g'
+	find $(dir $@)/input/regression -mindepth 2 -maxdepth 2 -name "*html" \
+	  | xargs $(PERL) -i -pe 's#(href)="css/#\1="../../../Documentation/css/#g'
 
 doc-build: top-doc
 	$(MAKE) out=www -C Documentation/ doc

Re: Error with make doc

[Carl Sorensen]

On Wed, Jan 31, 2024 at 12:49 AM Werner LEMBERG  wrote:

>
> > I already had to install (via homebrew) gnu coreutils and gnu
> > findutils, since the MacOS 'cp' and 'find' commands don't work the
> > same as the respective GNU commands.
>
> Please give more details.  What problems did you experience with `cp`
> and `find`?  Did this happen during a call to `make ...` for LilyPond?
>

All errors happened during 'make doc'.  They appeared to happen after all
of the lilypond processing was done, while creating the html documentation
pages.

I've attached three compressed log files.  One shows the cp error (when I
was using MacOS cp command by default).  The second shows the find error
(after configuring my system to use GNU cp).  The third shows the perl
error (after configuring my system to use GNU find as well as GNU cp).

This link shows an example of what might cause the cp error (as well as the
homebrew way to avoid it):
https://github.com/sahib/rmlint/issues/275

This link shows how to get the GNU find utilities on MacOS, but doesn't
reference the particular error in the log file.  It does, however, indicate
that there are a number of differences between the MacOS find and the GNU
find.:
https://unix.stackexchange.com/questions/277914/find-not-working

This link shows how the -name argument can lead to the observed error
message:
https://unix.stackexchange.com/questions/369771/find-command-not-working-illegal-option-n




> This should work out of the box, which essentially means that we have
> to either add tests to `configure` to catch the problematic tools, or
> to rewrite the commands in the makefile or scripts to avoid issues
> with GNU-style usage of those tools, for example.
>
> > I then get to a problem with perl: [...]
>
> Interesting.  I wasn't aware that we ever call `perl` with a long list
> of arguments during `make doc`.
>
> Please repeat this with saying
>
> ```
> make doc VERBOSE=1 &> make.doc.log
> ```
>
> so that we can more information in the log file.  Usually, such errors
> can easily be circumvented by using `xargs`.
>

That's the recommendation I found on the web, but I don't know the Makefile
well enough to propose a patch using xargs.

Thanks for all your work on this!

Carl


cp_error.make.doc.log.gz
Description: GNU Zip compressed data


find_error.make.doc.log.gz
Description: GNU Zip compressed data


perl_error.make.doc.log.gz
Description: GNU Zip compressed data

Re: Error with make doc

[Werner LEMBERG]

> I already had to install (via homebrew) gnu coreutils and gnu
> findutils, since the MacOS 'cp' and 'find' commands don't work the
> same as the respective GNU commands.

Please give more details.  What problems did you experience with `cp`
and `find`?  Did this happen during a call to `make ...` for LilyPond?
This should work out of the box, which essentially means that we have
to either add tests to `configure` to catch the problematic tools, or
to rewrite the commands in the makefile or scripts to avoid issues
with GNU-style usage of those tools, for example.

> I then get to a problem with perl: [...]

Interesting.  I wasn't aware that we ever call `perl` with a long list
of arguments during `make doc`.

Please repeat this with saying

```
make doc VERBOSE=1 &> make.doc.log
```

so that we can more information in the log file.  Usually, such errors
can easily be circumvented by using `xargs`.


Werner

Re: Error with make doc

[Carl Sorensen]

Still trying to get 'make doc' to complete successfully.

At this point, I think I'm through all the lilypond stuff, and now I'm
stuck in shell script errors.

I already had to install (via homebrew) gnu coreutils and gnu findutils,
since the MacOS 'cp' and 'find' commands don't work the same as the
respective GNU commands.

I then get to a problem with perl:

carl@carls-mbp-2 build % make doc
Making out-www/offline-root/index.html
Processing HTML pages for offline target...
/bin/sh: /opt/local/bin/perl: Argument list too long
make[1]: *** [out-www/offline-root/index.html] Error 126
make[1]: *** Deleting file `out-www/offline-root/index.html'
make: *** [doc] Error 2

The internet suggests I should look at getconf ARG_MAX:

build % getconf ARG_MAX
1048576

This seems like it should not be a problem.

Any suggestions for debugging this?

Thanks,

Carl

Re: Error with make doc

[Werner LEMBERG]

>> Aah, this looks like a bug in `typography-demo.ly`.  Please try the
>> following line in `typography-demo.ly` instead of the three
>> original font lines.
>>
>> ```
>> property-defaults.fonts.serif = "Linux Libertine O, Noto Serif CJK JP, Noto 
>> Serif JP, serif"
>> ```
>
> I can confirm that this change allows 'make test' to complete
> successfully.

Thanks for testing.

  https://gitlab.com/lilypond/lilypond/-/merge_requests/2244


 Werner

Re: Error with make doc

[Carl Sorensen]

On Wed, Jan 24, 2024 at 11:49 PM Werner LEMBERG  wrote:

>
> >> I installed Noto Serif JP, and added it to the fontcofig cache.
> >> That got me through the previous file error.  But then I got
> >> stopped on another Japanese font (I suppose this might be a Linux
> >> Libertine font, but my fc-list command shows that I have them
> >> installed):
>
> Interesting.  This shouldn't happen.  The idea is that the set of
> fonts used for building the documentation is closed and well-defined.
> 'Noto Serif JP' is intended to cover all CJK testing stuff in
> LilyPond.
>
> > I was able to successfully complete 'make test' by commenting out the
> > following lines in  input/regression/typography-demo.ly:
> >
> >   % Font settings for Cyrillic and Hebrew
> >   % Linux Libertine fonts contain Cyrillic and Hebrew glyphs.
> >   %property-defaults.fonts.serif = "Linux Libertine O,serif"
> >   %property-defaults.fonts.sans = "Linux Biolinum O,sans-serif"
> >   %property-defaults.fonts.typewriter = "Linux Libertine Mono
> O,monospace"
>
> Aah, this looks like a bug in `typography-demo.ly`.  Please try the
> following line in `typography-demo.ly` instead of the three original
> font lines.
>
> ```
> property-defaults.fonts.serif = "Linux Libertine O, Noto Serif CJK JP,
> Noto Serif JP, serif"
> ```
>
>
I can confirm that this change allows 'make test' to complete successfully.

Thanks!

Carl

Re: Error with make doc

[Werner LEMBERG]

>> I installed Noto Serif JP, and added it to the fontcofig cache.
>> That got me through the previous file error.  But then I got
>> stopped on another Japanese font (I suppose this might be a Linux
>> Libertine font, but my fc-list command shows that I have them
>> installed):

Interesting.  This shouldn't happen.  The idea is that the set of
fonts used for building the documentation is closed and well-defined.
'Noto Serif JP' is intended to cover all CJK testing stuff in
LilyPond.

> I was able to successfully complete 'make test' by commenting out the
> following lines in  input/regression/typography-demo.ly:
> 
>   % Font settings for Cyrillic and Hebrew
>   % Linux Libertine fonts contain Cyrillic and Hebrew glyphs.
>   %property-defaults.fonts.serif = "Linux Libertine O,serif"
>   %property-defaults.fonts.sans = "Linux Biolinum O,sans-serif"
>   %property-defaults.fonts.typewriter = "Linux Libertine Mono O,monospace"

Aah, this looks like a bug in `typography-demo.ly`.  Please try the
following line in `typography-demo.ly` instead of the three original
font lines.

```
property-defaults.fonts.serif = "Linux Libertine O, Noto Serif CJK JP, Noto 
Serif JP, serif"
```


Werner

Re: Error with make doc

[Carl Sorensen]

On Tue, Jan 23, 2024 at 4:54 AM Carl Sorensen 
wrote:

>
>
> On Mon, Jan 22, 2024 at 8:37 PM Carl Sorensen 
> wrote:
>
>>
>>
>> On Mon, Jan 22, 2024 at 5:58 PM Jean Abou Samra 
>> wrote:
>>
>>> Le lundi 22 janvier 2024 à 16:38 -0700, Carl Sorensen a écrit :
>>> > > Looks like no error:
>>> > >
>>> > > carl@Carls-MBP-2 lilypond % build/out/bin/lilypond input/regression/
>>> pdf-copy-paste.ly
>>> > > GNU LilyPond 2.25.13 (running Guile 3.0)
>>> > > Processing `input/regression/pdf-copy-paste.ly'
>>> > > Parsing...
>>> > > Finding the ideal number of pages...
>>> > > Fitting music on 1 page...
>>> > > Drawing systems...
>>> > > Converting to `pdf-copy-paste.pdf'...
>>> > > Success: compilation successfully completed
>>>
>>>
>>> Hmm. What if you run
>>>
>>> LC_ALL=C build/out/bin/lilypond input/regression/pdf-copy-paste.ly
>>>
>>> ? IIRC, tests are compiled with LC_ALL=C.
>>>
>>
>> OK, I'll run this test.  But it looks to me (see below).like I should be
>> able to solve the problem by getting the Noto Serif JP font.  I'll look at
>> installing that font.
>>
>
> I installed Noto Serif JP, and added it to the fontcofig cache.  That got
> me through the previous file error.  But then I got stopped on another
> Japanese font (I suppose this might be a Linux Libertine font, but my
> fc-list command shows that I have them installed):
>
> carl@carls-mbp-2 build % cat out/lybook-testdb/1d/lily-2548dc6c.log
>
> Processing `1d/lily-2548dc6c.ly'
>
> Parsing...
>
> Renaming input to: `/Users/carl/Development/lilypond/input/regression/
> typography-demo.ly'
>
> Interpreting music...[8]
>
> Preprocessing graphical objects...
>
> Interpreting music...
>
> MIDI output to `1d/lily-2548dc6c.midi'...
>
> Finding the ideal number of pages...
>
> Fitting music on 1 page...
>
> Drawing systems...
>
> Writing header field `texidoc' to `1d/lily-2548dc6c.texidoc'...
>
> Layout output to `1d/lily-2548dc6c.eps'...ERROR: In procedure open-file:
> In procedure open-file: No such file or directory:
> "/System/Library/Fonts/??? ProN.ttc"
>
>
I was able to successfully complete 'make test' by commenting out the
following lines in  input/regression/typography-demo.ly:

  % Font settings for Cyrillic and Hebrew

  % Linux Libertine fonts contain Cyrillic and Hebrew glyphs.

  %property-defaults.fonts.serif = "Linux Libertine O,serif"

  %property-defaults.fonts.sans = "Linux Biolinum O,sans-serif"

  %property-defaults.fonts.typewriter = "Linux Libertine Mono O,monospace"


iI appears that if Lilypond needs to search for a  font to display a
unicode character, and that font has Japanese characters in the font name,
then the font file is unable to be opened (as shown above).  There's been
some discussion between Jean and Werner indicating that it may be due to a
locale problem.  I suspect that an issue needs to be created, but I'm not
sure how to do it. And the issue may only show up on MacOS, because people
running Linux don't seem to be having this problem.


I'd welcome suggestions about how to create an issue for this, but I've got
a workaround that allows me to succeed with 'make test'.


Thanks,


Car

Re: Error with make doc

[Werner LEMBERG]

> carl@Carls-MBP-2 lilypond % tidy --version
> 
> HTML Tidy for Mac OS X released on 31 October 2006 - Apple Inc. build 6141

This explains the failure, thanks: Your `tidy` version doesn't support
HTML 5.  I've submitted an MR to catch this.

  https://gitlab.com/lilypond/lilypond/-/merge_requests/2241

> I didn't see tidy listed as a prerequisite.

It is not a prerequisite; it is simply used if available, as an
additional aid to the developers.  As it turns out, the naive test for
checking the presence of the binary is not sufficient; we must also
check for a proper version.

Please be reminded that up to now *nobody* worked on building the
(complete) documentation on the Mac!  You are effectively entering new
territory (for which I'm very thankful), and there are dragons, as
you've already found out.


Werner

Re: Error with make doc

[Carl Sorensen]

On Mon, Jan 22, 2024 at 8:37 PM Carl Sorensen 
wrote:

>
>
> On Mon, Jan 22, 2024 at 5:58 PM Jean Abou Samra 
> wrote:
>
>> Le lundi 22 janvier 2024 à 16:38 -0700, Carl Sorensen a écrit :
>> > > Looks like no error:
>> > >
>> > > carl@Carls-MBP-2 lilypond % build/out/bin/lilypond input/regression/
>> pdf-copy-paste.ly
>> > > GNU LilyPond 2.25.13 (running Guile 3.0)
>> > > Processing `input/regression/pdf-copy-paste.ly'
>> > > Parsing...
>> > > Finding the ideal number of pages...
>> > > Fitting music on 1 page...
>> > > Drawing systems...
>> > > Converting to `pdf-copy-paste.pdf'...
>> > > Success: compilation successfully completed
>>
>>
>> Hmm. What if you run
>>
>> LC_ALL=C build/out/bin/lilypond input/regression/pdf-copy-paste.ly
>>
>> ? IIRC, tests are compiled with LC_ALL=C.
>>
>
> OK, I'll run this test.  But it looks to me (see below).like I should be
> able to solve the problem by getting the Noto Serif JP font.  I'll look at
> installing that font.
>

I installed Noto Serif JP, and added it to the fontcofig cache.  That got
me through the previous file error.  But then I got stopped on another
Japanese font (I suppose this might be a Linux Libertine font, but my
fc-list command shows that I have them installed):

carl@carls-mbp-2 build % cat out/lybook-testdb/1d/lily-2548dc6c.log

Processing `1d/lily-2548dc6c.ly'

Parsing...

Renaming input to: `/Users/carl/Development/lilypond/input/regression/
typography-demo.ly'

Interpreting music...[8]

Preprocessing graphical objects...

Interpreting music...

MIDI output to `1d/lily-2548dc6c.midi'...

Finding the ideal number of pages...

Fitting music on 1 page...

Drawing systems...

Writing header field `texidoc' to `1d/lily-2548dc6c.texidoc'...

Layout output to `1d/lily-2548dc6c.eps'...ERROR: In procedure open-file:
In procedure open-file: No such file or directory:
"/System/Library/Fonts/??? ProN.ttc"

WHen I manually run lilypond on this file it succeeds:

carl@carls-mbp-2 build % out/bin/lilypond out/lybook-testdb/1d/
lily-2548dc6c.ly

GNU LilyPond 2.25.13 (running Guile 3.0)

Processing `out/lybook-testdb/1d/lily-2548dc6c.ly'

Parsing...

Renaming input to: `/Users/carl/Development/lilypond/input/regression/
typography-demo.ly'

Interpreting music...[8]

Preprocessing graphical objects...

Interpreting music...

MIDI output to `lily-2548dc6c.midi'...

Finding the ideal number of pages...

Fitting music on 1 page...

Drawing systems...

Converting to `lily-2548dc6c.pdf'...

Success: compilation successfully completed


It seems the difference may be I'm not creating the .eps file when I run
out/bin/lilypond.


What command could I run to try to duplicate the error in the log file?


Thanks,


Carl

Re: Error with make doc

[Werner LEMBERG]

>> Can someone remind me why the build system sets the C locale in the
>> first place?
> 
> Answering my own question: I guess it is because we can't assume the
> system has an en_US.UTF-8 locale or a C.UTF-8 locale?

Exactly.  Some time ago (see
https://lists.gnu.org/archive/html/lilypond-devel/2022-11/msg00247.html
and the following mails in the thread) I worked on fixing exactly that
(see attached old patch), but I stopped since we could then – at least
partially – circumvent the related problems.

I also stopped working on a `local-c-utf8.m4` file (also attached) to
recognize the 'C.UTF-8' and/or 'en_US.UTF-8' locales, which turned out
to be a really hard problem.

> Maybe we ought to set only LC_MESSAGES to C? It should prevent
> overriding the system locale for file names but still make logs
> output in English.

Honestly, I don't know.  The whole locale stuff is such a mess.


 Werner

From 231f5c2fba0fccf333fc7873da771fc8493b71e2 Mon Sep 17 00:00:00 2001
From: Werner Lemberg 
Date: Mon, 7 Nov 2022 09:38:17 +0100
Subject: [PATCH 1/2] generic-vars.make: Fix locale for build process
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

* In non-English indices of the PDF documentation, this fixes the initials
  for entries starting with a non-ASCII character.  This is a generic
  problem with the `texindex` awk script (used by the Texinfo infrastructure
  to sort indices).

https://lists.gnu.org/archive/html/bug-texinfo/2022-11/msg8.html

* Some non-ASCII characters are now correctly displayed in info, PDF, and
  HTML output files (example: `rotated 45?` → `rotated 45°` in the
  documentation of the `\rotate` markup).

Also update documentation accordingly.
---
 Documentation/en/contributor/lsr-work.itexi | 15 -
 Documentation/en/usage/running.itely|  2 +-
 Documentation/es/usage/running.itely|  3 ++-
 Documentation/fr/usage/running.itely|  2 +-
 Documentation/it/usage/running.itely|  2 +-
 Documentation/ja/usage/running.itely|  2 +-
 make/generic-vars.make  | 25 ++---
 python/book_latex.py|  2 +-
 python/book_texinfo.py  |  2 +-
 9 files changed, 39 insertions(+), 16 deletions(-)

diff --git a/Documentation/en/contributor/lsr-work.itexi b/Documentation/en/contributor/lsr-work.itexi
index b03ebf8122..c69d8ca485 100644
--- a/Documentation/en/contributor/lsr-work.itexi
+++ b/Documentation/en/contributor/lsr-work.itexi
@@ -445,12 +445,15 @@ updating the binary running the LSR.
 
 @item
 Download the latest snippet tarball from
-@uref{https://lsr.di.unimi.it/download/} and extract it.
-The relevant files can be found in the @file{all} subdirectory.
-Make sure your shell is using an English language version, for
-example @code{LANG=en_US}, then run @command{convert-ly} on all
-the files.  Use the command-line option @code{--to=version} to
-ensure the snippets are updated to the correct stable version.
+@uref{https://lsr.di.unimi.it/download/} and extract it; the
+relevant files can be found in the @file{all} subdirectory.
+
+Now run @code{LANG=C.UTF-8 convert-ly} on all the files.  The
+@code{LANG=C.UTF-8} part sets the @code{LANG} environment variable
+to a locale with UTF-8 encoding and disabled translations while
+@command{convert-ly} gets executed.  Also use the command-line
+option @code{--to=@var{version}} to ensure the snippets are
+updated to the correct stable version.
 
 @item
 Make sure that you are using @command{convert-ly} from the latest
diff --git a/Documentation/en/usage/running.itely b/Documentation/en/usage/running.itely
index a24c7ac392..029d1ae446 100644
--- a/Documentation/en/usage/running.itely
+++ b/Documentation/en/usage/running.itely
@@ -925,7 +925,7 @@ overrides the value derived from the location of the
 @item LANG
 The language for LilyPond data sent to @code{stdout} and
 @code{stderr}, for example progress reports, warning messages, or
-debug output.  Example: @code{LANG=de}.
+debug output.  Example: @samp{LANG=de_DE.UTF-8}.
 
 @item LILYPOND_LOGLEVEL
 The default loglevel.  If LilyPond is called without an explicit
diff --git a/Documentation/es/usage/running.itely b/Documentation/es/usage/running.itely
index 7f2629cd7d..7c722c50af 100644
--- a/Documentation/es/usage/running.itely
+++ b/Documentation/es/usage/running.itely
@@ -977,7 +977,8 @@ ubicación del binario @command{lilypond}.
 @item LANG
 Idioma de los datos de LilyPond enviados a @code{stdout} y a
 @code{stderr}, por ejemplo informes del avance, mensajes de
-advertencia o salida de depuración.  Ejemplo: @code{LANG=es}.
+advertencia o salida de depuración.  Ejemplo:
+@samp{LANG=es_ES.UTF-8}.
 
 @item LILYPOND_LOGLEVEL
 Nivel de registro predeterminado.  Si LilyPond se llama sin ningún
diff --git a/Documentation/fr/usage/running.itely b/Documentation/fr/usage/running.itely
index 5295ab0638..752ace6a2a 100644
--- 

Re: Error with make doc

[Carl Sorensen]

On Mon, Jan 22, 2024 at 5:58 PM Jean Abou Samra  wrote:

> Le lundi 22 janvier 2024 à 16:38 -0700, Carl Sorensen a écrit :
> > > Looks like no error:
> > >
> > > carl@Carls-MBP-2 lilypond % build/out/bin/lilypond input/regression/
> pdf-copy-paste.ly
> > > GNU LilyPond 2.25.13 (running Guile 3.0)
> > > Processing `input/regression/pdf-copy-paste.ly'
> > > Parsing...
> > > Finding the ideal number of pages...
> > > Fitting music on 1 page...
> > > Drawing systems...
> > > Converting to `pdf-copy-paste.pdf'...
> > > Success: compilation successfully completed
>
>
> Hmm. What if you run
>
> LC_ALL=C build/out/bin/lilypond input/regression/pdf-copy-paste.ly
>
> ? IIRC, tests are compiled with LC_ALL=C.
>

OK, I'll run this test.  But it looks to me (see below).like I should be
able to solve the problem by getting the Noto Serif JP font.  I'll look at
installing that font.

Here's the results of the test:

carl@carls-mbp-2 lilypond % LC_ALL=C build/out/bin/lilypond
input/regression/pdf-copy-paste.ly

GNU LilyPond 2.25.13 (running Guile 3.0)

Processing `input/regression/pdf-copy-paste.ly'

Parsing...

Finding the ideal number of pages...

Fitting music on 1 page...

Drawing systems...

Converting to `pdf-copy-paste.pdf'...

Success: compilation successfully completed

carl@carls-mbp-2 lilypond %


  As you can see, no error.


I'll work on the Noto JP Sans.

THanks,

Carl

Re: Error with make doc

[Jean Abou Samra]

> Can someone remind me why the build system sets the C locale
> in the first place?


Answering my own question: I guess it is because we can't assume the
system has an en_US.UTF-8 locale or a C.UTF-8 locale?

Maybe we ought to set only LC_MESSAGES to C? It should prevent overriding
the system locale for file names but still make logs output in English.



signature.asc
Description: This is a digitally signed message part

Re: Error with make doc

[Jean Abou Samra]

Ah, but, wait.

$ touch àéù.txt


$ guile3.0
scheme@(guile-user)> (open-file "àéù.txt" "r")
$1 = #

~/tmp $ LC_ALL=C guile3.0
scheme@(guile-user)> (open-file "àéù.txt" "r")
ice-9/boot-9.scm:1685:16: In procedure raise-exception:
In procedure open-file: No such file or directory: "??.txt"



OK, so Guile is assuming that the locale encoding determines
the encoding of file names (which is quite reasonable), and so,
when run under the C locale, it's transforming non-ASCII
characters into question marks before resolving the path,
as part of converting to the locale encoding.

Can someone remind me why the build system sets the C locale
in the first place?


signature.asc
Description: This is a digitally signed message part

Re: Error with make doc

[Jean Abou Samra]

Le lundi 22 janvier 2024 à 18:00 -0700, Carl Sorensen a écrit :
> So I ran the file in this log message:
> 
> 
> carl@carls-mbp-2 build % out/bin/lilypond out/share/lilypond/current/ly/
> init.ly
> 
> GNU LilyPond 2.25.13 (running Guile 3.0)
> 
> Processing `out/share/lilypond/current/ly/init.ly'
> 
> Parsing...
> 
> out/share/lilypond/current/ly/init.ly:47:1: error: \maininput not allowed
> outside init files
> 
> 
> \maininput
> 
> out/share/lilypond/current/ly/init.ly:1: warning: no \version statement
> found, please add
> 
> 
> \version "2.25.13"
> 
> 
> for future compatibility
> 
> out/share/lilypond/current/ly/init.ly:1: warning: no \version statement
> found, please add
> 
> 
> \version "2.25.13"
> 
> 
> for future compatibility
> 
> fatal error: failed files: "out/share/lilypond/current/ly/init.ly"
> 
> 
> Now I have found an error in the file -- it appears ist is due to
> \maininput not being allowed.


This isn't the problem. init.ly is a special file that LilyPond always
loads before starting to compile any file and is not meant to be
compiled standalone like this.

The reason the error message mentions init.ly is that init.ly
contains some Scheme code that is used to trigger the compilation.
(This could be improved.)

The problem really is the font issue in the error message.


signature.asc
Description: This is a digitally signed message part

Re: Error with make doc

[Jean Abou Samra]

Le mardi 23 janvier 2024 à 01:58 +0100, Jean Abou Samra a écrit :
> The latter is basically that this regtest (pdf-copy-paste.ly)
> has Japanese text, specifying the Noto Serif JP font, but since
> you apparently don't have that font installed, Fontconfig falls
> back to some system font, and apparently there is an encoding
> issue somewhere with the file name of that font but I don't know
> where exactly. Like Werner, I did some tests by renaming fonts
> on my machine, and there were no issues, even with LC_ALL=C.

(NB: The "??" in the error message is not indicative of a
problem per se. If the environment configures LC_ALL=C then
Guile detects that the current locale is not Unicode-aware
and falls back to printing non-ASCII characters as "?".
This doesn't necessarily mean that the string object being
printed actually has these question marks, it could be perfectly
correct.)


signature.asc
Description: This is a digitally signed message part

Re: Error with make doc

[Carl Sorensen]

On Mon, Jan 22, 2024 at 4:37 PM Carl Sorensen 
wrote:

>
>
> On Sun, Jan 21, 2024 at 12:33 AM Werner LEMBERG  wrote:
>
>>
>> >> You have to check the log files in `lybook-db` to find the problematic
>> >> `.ly` file(s).  Does `make test` succeed?  You should actually try that
>> >> before running `make doc`...
>> >>
>> >
>> > Interesting -- the CG has make doc show up before make test.  But I
>> > appreciate the suggestion, so I decided to restart
>> >
>> > So I did a make clean, and then a make test.
>>
>> I've never seen the errors you are reporting – as Dan writes, it may
>> be an issue with the `tidy` program.  However, I'm using version 5.4.0
>> on my GNU/Linux box while MacPorts offers the more recent version
>> 5.8.0; maybe there is indeed a problem in the created HTML code that
>> wasn't detected or handled by tidy in earlier versions.
>>
>> Another possibility is that macOS comes with an ancient `tidy` version
>> that doesn't understand HTML 5 stuff like ``, and you don't
>> have installed `tidy` via MacPorts at all.  In this case we have to
>> update the `output-distance` script to check the `tidy` version before
>> actually using it...
>>
>
> That looks like an issue.Apparently my 'tidy' is not installed by
> MacPorts.
>
> carl@carls-mbp-2 ly % which tidy
> /usr/bin/tidy
>
> and
>
> carl@Carls-MBP-2 lilypond % tidy --version
>
> HTML Tidy for Mac OS X released on 31 October 2006 - Apple Inc. build 6141
>
>
>
> I didn't see tidy listed as a prerequisite.
>
> After installing tidy in MacPorts, I get
>
> carl@carls-mbp-2 ~ % which tidy
>
> /opt/local/bin/tidy
>
>
> and
>
>
> carl@carls-mbp-2 ~ % tidy --version
>
> HTML Tidy for Apple macOS version 5.8.0
>
>
> I'll try rerunning 'make test' and see where I end up.  I think that you
> have clearly helped me identify a problem with tidy.
>
>
OK.  I deleted the build directory, and redid autogen.sh and ../configure.
And I also did make dist-clean (because I was asked to do so).

../configure succeeded.

'make all' succeeded.

'make test' gave this error:

Processing /Users/carl/Development/lilypond/build/out/lybook-testdb/
snippet-names-ece0c512da6daf7d9aff45809ce97556.ly

Command '/Users/carl/Development/lilypond/build/out/bin/lilypond -I
/Users/carl/Development/lilypond/input/regression -dseparate-page-formats=ps
-dseparate-log-files -ddeterministic -dinclude-eps-fonts -dgs-load-fonts
--header=texidoc -dcheck-internal-types -danti-alias-factor=1  -I
"/Users/carl/Development/lilypond/input/regression"  -I
"/Users/carl/Development/lilypond/build/input/regression"
-daux-files  -dread-file-list -dno-strip-output-dir -dtall-page-formats=eps
"/Users/carl/Development/lilypond/build/out/lybook-testdb/
snippet-names-ece0c512da6daf7d9aff45809ce97556.ly"' returned non-zero exit
status 1.

make[2]: *** [out-test/collated-files.texi] Error 1

make[1]: *** [lysdoc-test] Error 2

make: *** [test] Error 2


I tested the command that returned a non-zero exit status, and it succeeded.


carl@carls-mbp-2 build %
/Users/carl/Development/lilypond/build/out/bin/lilypond-I
/Users/carl/Development/lilypond/input/regression -dseparate-page-formats=ps
-dseparate-log-files -ddeterministic -dinclude-eps-fonts -dgs-load-fonts
--header=texidoc -dcheck-internal-types -danti-alias-factor=1  -I
"/Users/carl/Development/lilypond/input/regression"  -I
"/Users/carl/Development/lilypond/build/input/regression"
-daux-files  -dread-file-list -dno-strip-output-dir -dtall-page-formats=eps
"/Users/carl/Development/lilypond/build/out/lybook-testdb/
snippet-names-cce821d5aaa438c29a8c2ee16cd514b0.ly"

GNU LilyPond 2.25.13 (running Guile 3.0)

Processing `0f/lily-acefd898.ly'


I did the grep command suggested in the 'make test' output:


% grep -L systems.texi out/lybook-testdb/*/*log | sed s/log/ly/g | xargs
grep -H sourcefilename

out/lybook-testdb/0f/lily-acefd898.ly:\sourcefilename
"/Users/carl/Development/lilypond/input/regression/pdf-copy-paste.ly"


I ran the file identified in the grep output:


carl@carls-mbp-2 build % out/bin/lilypond ../input/regression/
pdf-copy-paste.ly

GNU LilyPond 2.25.13 (running Guile 3.0)

Processing
`/Users/carl/Development/lilypond/build/out/share/lilypond/current/ly/../input/regression/
pdf-copy-paste.ly'

Parsing...

Finding the ideal number of pages...

Fitting music on 1 page...

Drawing systems...

Converting to `pdf-copy-paste.pdf'...

Success: compilation successfully completed


I did the 'find' command suggested in the 'make test' output:

carl@carls-mbp-2 build % find out/lybook-testdb -name "*.log" | xargs grep
-H "\(error\|warning\)"

out/lybook-testdb/0f/lily-acefd898.log:/Users/carl/Development/lilypond/build/out/share/lilypond/current/ly/init.ly:66:2:
error: Guile signaled an error for the expression beginning here


So I ran the file in this log message:


carl@carls-mbp-2 build % out/bin/lilypond out/share/lilypond/current/ly/
init.ly

GNU LilyPond 2.25.13 (running Guile 3.0)

Processing `out/share/lilypond/current/ly/init.ly'

Re: Error with make doc

[Jean Abou Samra]

Le lundi 22 janvier 2024 à 16:38 -0700, Carl Sorensen a écrit :
> > Looks like no error:
> > 
> > carl@Carls-MBP-2 lilypond % build/out/bin/lilypond 
> > input/regression/pdf-copy-paste.ly
> > GNU LilyPond 2.25.13 (running Guile 3.0)
> > Processing `input/regression/pdf-copy-paste.ly'
> > Parsing...
> > Finding the ideal number of pages...
> > Fitting music on 1 page...
> > Drawing systems...
> > Converting to `pdf-copy-paste.pdf'...
> > Success: compilation successfully completed 


Hmm. What if you run

LC_ALL=C build/out/bin/lilypond input/regression/pdf-copy-paste.ly

? IIRC, tests are compiled with LC_ALL=C.


> > On  Werner's earlier email, he indicated there may be a problem with tidy.
> > I've gone ahead and installed tidy from MacPorts, and will try redoing
> > 'make test' again.


You have a problem with tidy *and* a problem with your fonts.

The latter is basically that this regtest (pdf-copy-paste.ly)
has Japanese text, specifying the Noto Serif JP font, but since
you apparently don't have that font installed, Fontconfig falls
back to some system font, and apparently there is an encoding
issue somewhere with the file name of that font but I don't know
where exactly. Like Werner, I did some tests by renaming fonts
on my machine, and there were no issues, even with LC_ALL=C.



signature.asc
Description: This is a digitally signed message part

Re: Error with make doc

[Carl Sorensen]

On Mon, Jan 22, 2024 at 3:58 PM Jean Abou Samra  wrote:

> Le lundi 22 janvier 2024 à 21:10 +, Werner LEMBERG a écrit :
> > > >
> > OK, thanks.  This means that there is apparently a macOS-specific
> > problem with either Guile or LilyPond.  For testing purposes I renamed
> > one of my local fonts to something containing Katakana, and my
> > self-compiled LilyPond binary could load this font without any glitch.
> >
> > Jean, Jonas, any idea how to proceed?
>
>
> Carl, please check that
>
> build/out/bin/lilypond input/regression/pdf-copy-paste.ly


Looks like no error:

carl@Carls-MBP-2 lilypond % build/out/bin/lilypond input/regression/
pdf-copy-paste.ly

GNU LilyPond 2.25.13 (running Guile 3.0)

Processing `input/regression/pdf-copy-paste.ly'

Parsing...

Finding the ideal number of pages...

Fitting music on 1 page...

Drawing systems...

Converting to `pdf-copy-paste.pdf'...
Success: compilation successfully completed

So it appears to me that there may be something with texidoc that is
generating the error.

On  Werner's earlier email, he indicated there may be a problem with tidy.
I've gone ahead and installed tidy from MacPorts, and will try redoing
'make test' again.

>
>
> gives the same error, then run
>
> build/out/bin/lilypond -ddebug-eval input/regression/pdf-copy-paste.ly
>
> so we have a more precise error location.
>
> Also send the output of the `locale` command.
>

carl@carls-mbp-2 Fonts % locale

LANG="en_US.UTF-8"

LC_COLLATE="en_US.UTF-8"

LC_CTYPE="en_US.UTF-8"

LC_MESSAGES="en_US.UTF-8"

LC_MONETARY="en_US.UTF-8"

LC_NUMERIC="en_US.UTF-8"

LC_TIME="en_US.UTF-8"
LC_ALL=

Thanks!

Carl

Re: Error with make doc

[Carl Sorensen]

On Sun, Jan 21, 2024 at 12:33 AM Werner LEMBERG  wrote:

>
> >> You have to check the log files in `lybook-db` to find the problematic
> >> `.ly` file(s).  Does `make test` succeed?  You should actually try that
> >> before running `make doc`...
> >>
> >
> > Interesting -- the CG has make doc show up before make test.  But I
> > appreciate the suggestion, so I decided to restart
> >
> > So I did a make clean, and then a make test.
>
> I've never seen the errors you are reporting – as Dan writes, it may
> be an issue with the `tidy` program.  However, I'm using version 5.4.0
> on my GNU/Linux box while MacPorts offers the more recent version
> 5.8.0; maybe there is indeed a problem in the created HTML code that
> wasn't detected or handled by tidy in earlier versions.
>
> Another possibility is that macOS comes with an ancient `tidy` version
> that doesn't understand HTML 5 stuff like ``, and you don't
> have installed `tidy` via MacPorts at all.  In this case we have to
> update the `output-distance` script to check the `tidy` version before
> actually using it...
>

That looks like an issue.Apparently my 'tidy' is not installed by
MacPorts.

carl@carls-mbp-2 ly % which tidy
/usr/bin/tidy

and

carl@Carls-MBP-2 lilypond % tidy --version

HTML Tidy for Mac OS X released on 31 October 2006 - Apple Inc. build 6141



I didn't see tidy listed as a prerequisite.

After installing tidy in MacPorts, I get

carl@carls-mbp-2 ~ % which tidy

/opt/local/bin/tidy


and


carl@carls-mbp-2 ~ % tidy --version

HTML Tidy for Apple macOS version 5.8.0


I'll try rerunning 'make test' and see where I end up.  I think that you
have clearly helped me identify a problem with tidy.


Thanks,


Carl

Re: Error with make doc

[Jean Abou Samra]

Le lundi 22 janvier 2024 à 21:10 +, Werner LEMBERG a écrit :
> > > 
> OK, thanks.  This means that there is apparently a macOS-specific
> problem with either Guile or LilyPond.  For testing purposes I renamed
> one of my local fonts to something containing Katakana, and my
> self-compiled LilyPond binary could load this font without any glitch.
> 
> Jean, Jonas, any idea how to proceed?


Carl, please check that

build/out/bin/lilypond input/regression/pdf-copy-paste.ly

gives the same error, then run

build/out/bin/lilypond -ddebug-eval input/regression/pdf-copy-paste.ly

so we have a more precise error location.

Also send the output of the `locale` command.

Thanks.



signature.asc
Description: This is a digitally signed message part

Re: Error with make doc

[Werner LEMBERG]

>> > In procedure open-file: No such file or directory:
>> > "/System/Library/Fonts/??? W4.ttc" [...]
>> >
>> > SO it looks to me like there's a problem with not getting UTF characters
>> > into the proper file name in the Scheme procedure.  Any ideas for what to
>> > do next?
>>
>> What does `fc-list` report?  Do you see correct file names?
>> (`fc-list` should come with the FontConfig library).
>
> Shows the fonts with the appropriate kana/kanji names:
>
> carl@carls-mbp-2 Fonts % fc-list | grep W4
>
> /System/Library/Fonts/ヒラギノ丸ゴ ProN W4.ttc: Hiragino Maru Gothic ProN,
> [...]

OK, thanks.  This means that there is apparently a macOS-specific
problem with either Guile or LilyPond.  For testing purposes I renamed
one of my local fonts to something containing Katakana, and my
self-compiled LilyPond binary could load this font without any glitch.

Jean, Jonas, any idea how to proceed?


Werner

Re: Error with make doc

[Carl Sorensen]

On Sun, Jan 21, 2024 at 11:05 PM Werner LEMBERG  wrote:

> > In procedure open-file: No such file or directory:
> > "/System/Library/Fonts/??? W4.ttc"
> >
> > I checked out the /System/Library/Fonts directory, and found these fonts
> > listed:
> >
> > ヒラギノ角ゴシック W3.ttc
> >
> > ヒラギノ角ゴシック W4.ttc
> >
> > SO it looks to me like there's a problem with not getting UTF characters
> > into the proper file name in the Scheme procedure.  Any ideas for what to
> > do next?
>
> What does `fc-list` report?  Do you see correct file names?
> (`fc-list` should come with the FontConfig library).
>

Shows the fonts with the appropriate kana/kanji names:

carl@carls-mbp-2 Fonts % fc-list | grep W4

/System/Library/Fonts/ヒラギノ丸ゴ ProN W4.ttc: Hiragino Maru Gothic
ProN,ヒラギノ丸ゴ ProN,Hiragino Maru Gothic ProN W4,ヒラギノ丸ゴ ProN
W4:style=W4,Regular

/System/Library/Fonts/ヒラギノ角ゴシック W4.ttc: Hiragino Sans,ヒラギノ角ゴシック,Hiragino
Sans W4,ヒラギノ角ゴシック W4:style=W4,Regular

/System/Library/Fonts/ヒラギノ角ゴシック W4.ttc: .Hiragino Kaku Gothic
Interface:style=W4
/System/Library/Fonts/ヒラギノ丸ゴ ProN W4.ttc: Hiragino Maru Gothic Pro,ヒラギノ丸ゴ
Pro,Hiragino Maru Gothic Pro W4,ヒラギノ丸ゴ Pro W4:style=W4,Regular

Thanks,

Carl

Re: Error with make doc

[Werner LEMBERG]

> In procedure open-file: No such file or directory:
> "/System/Library/Fonts/??? W4.ttc"
>
> I checked out the /System/Library/Fonts directory, and found these fonts
> listed:
>
> ヒラギノ角ゴシック W3.ttc
>
> ヒラギノ角ゴシック W4.ttc
>
> SO it looks to me like there's a problem with not getting UTF characters
> into the proper file name in the Scheme procedure.  Any ideas for what to
> do next?

What does `fc-list` report?  Do you see correct file names?
(`fc-list` should come with the FontConfig library).


Werner

Re: Error with make doc

[Carl Sorensen]

Thanks for the hint, Dan!

I got to a new log file that gave me what I believe is a helpful error
message:

carl@carls-mbp-2 build % cat out/lybook-testdb/0f/lily-acefd898.log

Processing `0f/lily-acefd898.ly'

Parsing...

Renaming input to: `/Users/carl/Development/lilypond/input/regression/
pdf-copy-paste.ly'

Writing header field `texidoc' to `0f/lily-acefd898.texidoc'...

Layout output to `0f/lily-acefd898.eps'...

/Users/carl/Development/lilypond/build/out/share/lilypond/current/ly/init.ly:66:2:
error: Guile signaled an error for the expression beginning here

#

 (let ((book-handler (if (defined? 'default-toplevel-book-handler)

In procedure open-file: No such file or directory:
"/System/Library/Fonts/??? W4.ttc"


I checked out the /System/Library/Fonts directory, and found these fonts
listed:


ヒラギノ角ゴシック W3.ttc

ヒラギノ角ゴシック W4.ttc


SO it looks to me like there's a problem with not getting UTF characters
into the proper file name in the Scheme procedure.  Any ideas for what to
do next?


Thanks,


Carl

>
>

Re: Error with make doc

[Werner LEMBERG]

>> You have to check the log files in `lybook-db` to find the problematic
>> `.ly` file(s).  Does `make test` succeed?  You should actually try that
>> before running `make doc`...
>>
> 
> Interesting -- the CG has make doc show up before make test.  But I
> appreciate the suggestion, so I decided to restart
> 
> So I did a make clean, and then a make test.

I've never seen the errors you are reporting – as Dan writes, it may
be an issue with the `tidy` program.  However, I'm using version 5.4.0
on my GNU/Linux box while MacPorts offers the more recent version
5.8.0; maybe there is indeed a problem in the created HTML code that
wasn't detected or handled by tidy in earlier versions.

Another possibility is that macOS comes with an ancient `tidy` version
that doesn't understand HTML 5 stuff like ``, and you don't
have installed `tidy` via MacPorts at all.  In this case we have to
update the `output-distance` script to check the `tidy` version before
actually using it...


Werner

Re: Error with make doc

[Dan Eble]

On 2024-01-20 20:55, Carl Sorensen wrote:


   File
"/Library/Frameworks/Python.framework/Versions/3.10/lib/python3.10/subprocess.py",
line 524, in run

 raise CalledProcessError(retcode, process.args,

subprocess.CalledProcessError: Command '['/usr/bin/tidy', '-o',
'/dev/null', '-q', 'compare-dir1dir2/index.html']' returned non-zero exit
status 2.


Try commenting out this part of scripts/build/output-distance.py and 
retrying:


    tidy_bin = shutil.which('tidy')
    if tidy_bin:
    subprocess.run([tidy_bin, '-o', '/dev/null', '-q', html_fn], 
check=True)


I'm not sure that that's the only problem, but it should eliminate one 
thing.


--
Dan

Re: Error with make doc

[Carl Sorensen]

On Thu, Jan 18, 2024 at 11:05 PM Werner LEMBERG  wrote:

>
> > "/Users/carl/Development/lilypond/build/out/lybook-db/
> > snippet-names-ece0c512da6daf7d9aff45809ce97556.ly"' returned non-zero
> exit
> > status 1.
>
> You have to check the log files in `lybook-db` to find the problematic
> `.ly` file(s).  Does `make test` succeed?  You should actually try that
> before running `make doc`...
>

Interesting -- the CG has make doc show up before make test.  But I
appreciate the suggestion, so I decided to restart

So I did a make clean, and then a make test.

Got an error, and this is the log file  (
/my_path/lilypond/build/scripts/build/./out/output-distance-test.log ):

Command: /Library/Frameworks/Python.framework/Versions/3.10/bin/python3 -W
default -W ignore::ResourceWarning
/Users/carl/Development/lilypond/scripts/build/output-distance.py --test
--output-dir=./out


GNU LilyPond 2.25.13 (running Guile 3.0)

Processing `20'


Processing `19'


Processing `20grob'


Processing `20multipage'

Processing `19multipage'

Processing `added'


Processing `removed'


Processing `added-multipage.ly'


Success: compilation successfully completed

line 96 column 7 - Error:  is not recognized!

line 96 column 7 - Warning: discarding unexpected 

line 97 column 1 - Error:  is not recognized!

line 97 column 1 - Warning: discarding unexpected 

line 99 column 1 - Warning: discarding unexpected 

line 100 column 1 - Error:  is not recognized!

line 100 column 1 - Warning: discarding unexpected 

line 100 column 57 - Warning: discarding unexpected 

line 101 column 1 - Warning: discarding unexpected 

line 109 column 7 - Error:  is not recognized!

line 109 column 7 - Warning: discarding unexpected 

line 110 column 1 - Error:  is not recognized!

line 110 column 1 - Warning: discarding unexpected 

line 112 column 1 - Warning: discarding unexpected 

line 114 column 1 - Error:  is not recognized!

test results in  ./out/test-output-distance

comparing dir1

   to dir2

25 eps

 2 midi

 7 log

 1 gittxt

35 total

no source for dir2/context.ly

no source for dir2/removed.ly

no source for dir2/midi-differ.midi

no source for dir2/log-differ.log

no source for dir2/removed.log

no source for dir2/tree.gittxt

converting 4 EPS files

converting 16 EPS files

converting 1 EPS files

output-distance summary:

13 changed

 3 in baseline only

 0 below threshold

 3 unchanged

Traceback (most recent call last):

  File "/my_path/lilypond/scripts/build/output-distance.py", line 1424, in


main()

  File "/my_path/lilypond/scripts/build/output-distance.py", line 1412, in
main

run_tests()

  File "/my_path/lilypond/scripts/build/output-distance.py", line 1359, in
run_tests

test_compare_tree_pairs()

  File "/my_path/lilypond/scripts/build/output-distance.py", line 1199, in
test_compare_tree_pairs

subprocess.run([tidy_bin, '-o', '/dev/null', '-q', html_fn], check=True)

  File
"/Library/Frameworks/Python.framework/Versions/3.10/lib/python3.10/subprocess.py",
line 524, in run

raise CalledProcessError(retcode, process.args,

subprocess.CalledProcessError: Command '['/usr/bin/tidy', '-o',
'/dev/null', '-q', 'compare-dir1dir2/index.html']' returned non-zero exit
status 2.

I'm not sure how to interpret or track down this error further.  Any help
would be appreciated.

Thanks,

Carl

Re: Error with make doc

[Werner LEMBERG]

> "/Users/carl/Development/lilypond/build/out/lybook-db/
> snippet-names-ece0c512da6daf7d9aff45809ce97556.ly"' returned non-zero exit
> status 1.

You have to check the log files in `lybook-db` to find the problematic
`.ly` file(s).  Does `make test` succeed?  You should actually try that
before running `make doc`...

BTW, if you execute `make test`, you are greeted with the following
advice:

```
To begin investigating regression-test crashes, try this:

grep -L systems.texi out/lybook-testdb/*/*log | sed s/log/ly/g | xargs grep 
-H sourcefilename

Matching lines may provide original test source file names.
The .ly files in which they appear are preprocessed test input.
The corresponding .log files may have errors or warnings.

If the above is not helpful, try sifting through this:

find out/lybook-testdb -name "*.log" | xargs grep -H "\(error\|warning\)"
```

which can also be used for finding problems with `make doc` (but you
have to replace `lybook-testdb` with `lybook-db` in the above
commands).


Werner

Re: Error from "make doc"

[James]

Federico,

On 22/05/16 21:37, Federico Bruni wrote:
Il giorno dom 22 mag 2016 alle 11:52, James Lowe  ha 
scritto:

 Install the package lmodern:

 sudo apt-get install lmodern

 Added to LilyDev only one month ago:

https://github.com/fedelibre/LilyDev/commit/2d4ed72d0d0a5f1b73b71d80489524c624e5727f 







I am confused here.

Had John not used LilyDev but followed the documented CG instructions 
for whatever Linux distribution - had he had one already installed - 
then he wouldn't have hit this issue I suspect, I say that as someone 
who compiles doc most days on both Ubuntu and Fedora and have tested 
and documented all the instructions for all the distributions listed 
in the CG and on my own 'at home' Fedora install I don't see any 
'modern' fonts installed if I run the command:


fc-list | grep modern

(or 'grep lm')

I can also see that the XeteX package listed in Debian's repos shows 
lmodern as 'recommended' however on my own system I have 
texlive-xetex installed and yet I still don't see lmodern installed, 
so it must not be a dep on Fedora either.


lmodern is not a req as far as I can tell.

So what is special about LilyDev?


I can only say that when I tried to build the doc with xetex installed 
(and lmdodern not installed) I got this error in LilyDev/Debian:


   For xetex backend Debian package lmodern is needed


In Fedora the package is called texlive-lm.
This command returns a list of the installed files:

rpm -ql texlive-lm

I can only guess that these fonts are supposed to be available only to 
Texlive and not to the whole system?!


Yes that seems correct when I look at my system I do have those fonts.

So it also seems that the texlive-xetex package and the various 
dependencies are different across one or more distributions - Linux eh? 
'Doncha juss lovvit?'


Going back through my checkins for the CG I also note that we didn't 
suggest XeTeX when I did all those different distributions, so we *do* 
need to make some additional caveats.


I'll create a ticket.

James

___
lilypond-devel mailing list
lilypond-devel@gnu.org
https://lists.gnu.org/mailman/listinfo/lilypond-devel

Re: Error from "make doc"

[Federico Bruni]

Il giorno dom 22 mag 2016 alle 11:52, James Lowe  ha 
scritto:

 Install the package lmodern:

 sudo apt-get install lmodern

 Added to LilyDev only one month ago:

https://github.com/fedelibre/LilyDev/commit/2d4ed72d0d0a5f1b73b71d80489524c624e5727f





I am confused here.

Had John not used LilyDev but followed the documented CG instructions 
for whatever Linux distribution - had he had one already installed - 
then he wouldn't have hit this issue I suspect, I say that as someone 
who compiles doc most days on both Ubuntu and Fedora and have tested 
and documented all the instructions for all the distributions listed 
in the CG and on my own 'at home' Fedora install I don't see any 
'modern' fonts installed if I run the command:


fc-list | grep modern

(or 'grep lm')

I can also see that the XeteX package listed in Debian's repos shows 
lmodern as 'recommended' however on my own system I have 
texlive-xetex installed and yet I still don't see lmodern installed, 
so it must not be a dep on Fedora either.


lmodern is not a req as far as I can tell.

So what is special about LilyDev?


I can only say that when I tried to build the doc with xetex installed 
(and lmdodern not installed) I got this error in LilyDev/Debian:


   For xetex backend Debian package lmodern is needed


In Fedora the package is called texlive-lm.
This command returns a list of the installed files:

rpm -ql texlive-lm

I can only guess that these fonts are supposed to be available only to 
Texlive and not to the whole system?!?





___
lilypond-devel mailing list
lilypond-devel@gnu.org
https://lists.gnu.org/mailman/listinfo/lilypond-devel

Re: Error from "make doc"

[James Lowe]

Federico,

On 22/05/16 08:43, Federico Bruni wrote:
I don't think so. It's supposed to work out-of-the-box in LilyDev, but 
I forgot to include this package in LilyDev 4.1. See warning in the 
release note:

https://github.com/fedelibre/LilyDev/releases/tag/4.1

Il giorno dom 22 mag 2016 alle 3:54, John Gourlay 
 ha scritto:
Thank you! Installing the package lmodern fixed the problem. 
Shouldn’t this instruction be part of the Contributor’s Guide?


 On May 20, 2016, at 1:36 AM, Federico Bruni  
wrote:


 Il giorno ven 20 mag 2016 alle 5:10, John Gourlay 
 ha scritto:
 I’m trying to set up a new LilyDev development system after some 
kind of VirtualBox error that trashed my old one. I’ve run “make” 
successfully in my ~/lilypond-git/build directory, and lilypond 
runs. When I run “make doc”, however, I get errors. One of the 
messages from make directed me to look at a log file named 
suffix-lyxml.dblatex.log. This file contains the following messages 
about invalid fonts. Have I overlooked something in setting up my 
system?


 Install the package lmodern:

 sudo apt-get install lmodern

 Added to LilyDev only one month ago:

https://github.com/fedelibre/LilyDev/commit/2d4ed72d0d0a5f1b73b71d80489524c624e5727f 







I am confused here.

Had John not used LilyDev but followed the documented CG instructions 
for whatever Linux distribution - had he had one already installed - 
then he wouldn't have hit this issue I suspect, I say that as someone 
who compiles doc most days on both Ubuntu and Fedora and have tested and 
documented all the instructions for all the distributions listed in the 
CG and on my own 'at home' Fedora install I don't see any 'modern' fonts 
installed if I run the command:


fc-list | grep modern

(or 'grep lm')

I can also see that the XeteX package listed in Debian's repos shows 
lmodern as 'recommended' however on my own system I have texlive-xetex 
installed and yet I still don't see lmodern installed, so it must not be 
a dep on Fedora either.


lmodern is not a req as far as I can tell.

So what is special about LilyDev?

--
--

James

___
lilypond-devel mailing list
lilypond-devel@gnu.org
https://lists.gnu.org/mailman/listinfo/lilypond-devel

Re: Error from "make doc"

[Federico Bruni]

I don't think so. It's supposed to work out-of-the-box in LilyDev, but 
I forgot to include this package in LilyDev 4.1. See warning in the 
release note:

https://github.com/fedelibre/LilyDev/releases/tag/4.1

Il giorno dom 22 mag 2016 alle 3:54, John Gourlay 
 ha scritto:
Thank you! Installing the package lmodern fixed the problem. 
Shouldn’t this instruction be part of the Contributor’s Guide?


 On May 20, 2016, at 1:36 AM, Federico Bruni  
wrote:


 Il giorno ven 20 mag 2016 alle 5:10, John Gourlay 
 ha scritto:
 I’m trying to set up a new LilyDev development system after some 
kind of VirtualBox error that trashed my old one. I’ve run 
“make” successfully in my ~/lilypond-git/build directory, and 
lilypond runs. When I run “make doc”, however, I get errors. 
One of the messages from make directed me to look at a log file 
named suffix-lyxml.dblatex.log. This file contains the following 
messages about invalid fonts. Have I overlooked something in 
setting up my system?


 Install the package lmodern:

 sudo apt-get install lmodern

 Added to LilyDev only one month ago:
 
https://github.com/fedelibre/LilyDev/commit/2d4ed72d0d0a5f1b73b71d80489524c624e5727f







___
lilypond-devel mailing list
lilypond-devel@gnu.org
https://lists.gnu.org/mailman/listinfo/lilypond-devel

Re: Error from "make doc"

[John Gourlay]

Thank you! Installing the package lmodern fixed the problem. Shouldn’t this 
instruction be part of the Contributor’s Guide?

> On May 20, 2016, at 1:36 AM, Federico Bruni  wrote:
> 
> Il giorno ven 20 mag 2016 alle 5:10, John Gourlay  
> ha scritto:
>> I’m trying to set up a new LilyDev development system after some kind of 
>> VirtualBox error that trashed my old one. I’ve run “make” successfully in my 
>> ~/lilypond-git/build directory, and lilypond runs. When I run “make doc”, 
>> however, I get errors. One of the messages from make directed me to look at 
>> a log file named suffix-lyxml.dblatex.log. This file contains the following 
>> messages about invalid fonts. Have I overlooked something in setting up my 
>> system?
> 
> Install the package lmodern:
> 
> sudo apt-get install lmodern
> 
> Added to LilyDev only one month ago:
> https://github.com/fedelibre/LilyDev/commit/2d4ed72d0d0a5f1b73b71d80489524c624e5727f
> 


___
lilypond-devel mailing list
lilypond-devel@gnu.org
https://lists.gnu.org/mailman/listinfo/lilypond-devel

Re: Error from "make doc"

[Federico Bruni]

Il giorno ven 20 mag 2016 alle 5:10, John Gourlay 
 ha scritto:
I’m trying to set up a new LilyDev development system after some 
kind of VirtualBox error that trashed my old one. I’ve run 
“make” successfully in my ~/lilypond-git/build directory, and 
lilypond runs. When I run “make doc”, however, I get errors. One 
of the messages from make directed me to look at a log file named 
suffix-lyxml.dblatex.log. This file contains the following messages 
about invalid fonts. Have I overlooked something in setting up my 
system?


Install the package lmodern:

sudo apt-get install lmodern

Added to LilyDev only one month ago:
https://github.com/fedelibre/LilyDev/commit/2d4ed72d0d0a5f1b73b71d80489524c624e5727f


___
lilypond-devel mailing list
lilypond-devel@gnu.org
https://lists.gnu.org/mailman/listinfo/lilypond-devel