Re: Output from woodwind-diagrams-key-lists.ly

[David Kastrup]

"Phil Holmes" <address@hidden> writes:

> Some of the output from make doc comes deliberately from regression
> tests. For example, woodwind-diagrams-key-lists.ly puts lots of
> information similar to the following on the terminal screen:
>
> Printing keys in verbose mode for: piccolocc   possibilities for one:
> (one oneT oneRT oneTR oneR oneRT1q one1qTR oneRT1h one1hTR oneRT3q
> one3qTR oneRTF oneFTR one1qT oneT1q one1q one1qT1h one1hT1q one1qT3q
> one3qT1q one1qTF oneFT1q one1hT oneT1h one1h one1hT3q one3qT1h one1hTF
> oneFT1h one3qT oneT3q one3q one3qTF oneFT3q oneFT oneF)
>
> Which is all very well, if it were ever checked.  Compiling same file
> in windows by dropping the file on the Lily icon produces a logfile
> with none of the above - it just vanishes.  I think it's highly likely
> that the same is true for the regtest logfiles.  Part of the reason
> for this would appear to be that the print-keys-verbose instruction
> prints to stdout, whereas pretty much all other lily output goes to
> stderr.
>
> I'd like to get rid of it.  Options would appear to be:
>
> 1. Make a special case for output like this and redirect the stdout as
> well as stderr output to a logfile - currently I've not done this
> since for all the other files, it adds no value.

I think that this is more or less done on Unix-like systems, and it is a
mess since stdout and stderr apparently get flushed asynchronously, so
the output order differs from run to run and does not always make sense.

> 2. Rewrite display-woodwind-diagrams.scm so that it outputs to stderr
> - this would make the regtest actually work, since the logfile would
> be compared

Things like print-keys etc should get an optional port argument
defaulting to the current output port.  The regtests should use this for
directing their output to where it is expected.

> 3. Get rid of the regtest.

Getting rid of regtests because they cause problems is always a bit
suspicious.

-- 
David Kastrup