Re: make doc

[Phil Holmes]

----- Original Message ----- 
From: "Graham Percival" <address@hidden>
To: "Phil Holmes" <address@hidden>
Cc: <address@hidden>
Sent: Wednesday, June 01, 2011 3:38 PM
Subject: Re: make doc


> On Wed, Jun 01, 2011 at 03:27:14PM +0100, Phil Holmes wrote:
> > ----- Original Message ----- From: "Graham Percival"
> > <address@hidden>
> > >If that's true, then maybe this should be added.  It'd be another
> > >of those 5-lines patches that takes 5 hours to write, though.
> >
> > I don't think so - to do this, I'd propose to add a flag to
> > lilypond-book, -r.  If the -r flag is present, l-book would redirect
> > all output from compiling 1234abcd.ly to 1234abcd.log in the same
> > directory as the input file.  You OK with that?
>
> That would be absolutely ideal!  I'm not totally sold on using -r
> since that tends to be used for "recursive".  How about just make
> it --redirect-lilypond-output ?  This isn't a command that people
> will need to type by hand a lot (or even), so I think a
> long+descriptive name would be perfect.


I've concluded that your "5 lines takes 5 hours" is close to the truth.  To 
make lilypond-book redirect lilypond output is non-trivial, but I now know 
how.  It means editing lilylib.py, but only to make an existing option 
functional.  Here's my notes about how this would be done.  I'm guessing 
your response will be - go ahead, submit a patch to Rietveld...

We need to edit lilypond-book.  The relevant line is:

system_in_directory (' '.join ([cmd, ly.mkarg (name)]),
                        lily_output_dir)

This concatenates the lilypond-run command with the filename to be
compiled and supplies this as the first arg to a function that
runs a system command, like this:

ly.system(cmd, be_verbose=global_options.verbose,
             progress_p=1)

The system command is found in lilylib.py:

system = subprocess_system
if sys.platform == 'mingw32':
   ## subprocess x-compile doesn't work.
   system = ossystem_system

and elsewhere in the file:

def subprocess_system (cmd,
                      ignore_error=False,
                      progress_p=True,
                      be_verbose=False,
                      log_file=None):

That's interesting - it already has a log_file parameter - which
isn't used....  Currently the non-windows version has this:

   stdout_setting = None
   if not show_progress:
       stdout_setting = subprocess.PIPE

   proc = subprocess.Popen (cmd,
       shell=True,
       universal_newlines=True,
       stdout=stdout_setting,
       stderr=stdout_setting)

If we change this to something like:

   out = open('stdout.txt', 'w')
   proc = subprocess.Popen (cmd,
       shell=True,
       universal_newlines=True,
       stdout=out,
       stderr=out)
   out.close()

we get the output sent to stdout.txt. :-) So all (!) we need to do
is use the log_file parameter (I suggest this should be a string,
not an open file, and to append .log in the opening process above)
to create a file and send the output to.  If we wanted to be
really neat, we could actually have 2 files -
lilyfile.log and lilyfile.error.log

--
Phil Holmes