#5780 Accept GUILE 2 without extra configure options

[Han-Wen Nienhuys]

A run of the regression tests at "5f4cf7bd38 - Parse inline scheme
using per-expression port" shows the following issues compared to
GUILE 1.8

• changed interleaving of error and progress messages.

• floating point rounding differences.

• an encoding error in song-basic-nonenglish.ly

As of a19aed1 "GUILE2: Scale GC heap with the number of smobs",
performance for large scores with GUILE 2.2 is similar to GUILE 1.8.

For small scores, a 1.5 sec extra delay (due to byte-code compilation
of .scm files) remains, but this does not warrant the "highly
experimental" monniker.

https://codereview.appspot.com/569390043

[Han-Wen Nienhuys]

guile-config after guile-config2.2

https://codereview.appspot.com/569390043

[Han-Wen Nienhuys]

pkg-config

https://codereview.appspot.com/569390043

[Han-Wen Nienhuys]

stats

https://codereview.appspot.com/569390043

[Han-Wen Nienhuys]

jonas' comments

https://codereview.appspot.com/569390043

[Anonymous]

Passes make, make check and a full make doc.

[Anonymous]

Patch on countdown for Feb 24th

[Anonymous]

Patch counted down - please push

[Jonas Hahnfeld]

commit 670e9f7b9660822f31144cbf07788bfc28de94cd
Author:     Han-Wen Nienhuys <hanwen@lilypond.org>
AuthorDate: Fri Feb 21 16:14:48 2020 +0100
Commit:     Han-Wen Nienhuys <hanwen@lilypond.org>
CommitDate: Mon Feb 24 21:50:15 2020 +0100

    Accept GUILE 2 without extra configure options

    A run of the regression tests at "5f4cf7bd38 - Parse inline scheme
    using per-expression port" shows the following issues compared to
    GUILE 1.8

    * changed interleaving of error and progress messages.

    * floating point rounding differences.

    * an encoding error in song-basic-nonenglish.ly

    As of a19aed1 "GUILE2: Scale GC heap with the number of smobs",
    performance for large scores with GUILE 2.2 is just ~25% slower on
    large scores.

    The regression test (make test-baseline) is 67% slower. This can
    partly be explained because of the slower startup: until we can
    support pre-compiled files, we incur a 1.5 sec extra delay for parsing
    .scm files.

[David Kastrup]

I think we need to redial or augment some of this patch since it will leave even experienced developers with an unexpected result.

If I now do
GUILE_CONFIG=/usr/local/tmp/guile-1.8/bin/guile-config ./configure
on my system the slow and crash-prone Guile-2.2 will be used. I would have been notified of that because the configuration would have failed because of a missing --with-guilev2 option, but this option is no longer necessary. So my explicit choice of Guile-1.8, made in the previously supported manner, will get silently ignored.

That is not good. I think that if GUILE_CONFIG is given, it should be heeded rather than ignored. Also if PKG_CONFIG_PATH is given in the environment, it should be recorded so that
./config.status --recheck
will not forget about it again.

We have dropped too many safety nets for making sure that our configure does what is expected of it.