Re: guile-2.0 and debian

[Thomas Morley]

2016-11-07 17:26 GMT+01:00 David Kastrup <address@hidden>:
> Federico Bruni <address@hidden> writes:
>
>> In case you don't know already, last news about guile2 and debian:
>> https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=746005#216
>>
>> As you can see, Antonio Ospite is trying to work on it and wrote some
>> notes:
>> https://ao2.it/tmp/lilypond-guile2/NOTES.txt
>>
>> Don't know if he already tried to contact any of you (David?)
>> privately...
>
> Don't remember who did it but I received a copy.  Most stuff in the
> notes is already fixed in master.
>
> --
> David Kastrup



Regrettable I was not able to do serious work on guilev2 so far.

Nevertheless, I read what Antonio Ospite did and tried a naive approach.
First changing my system to use guilev2 (which is "2.0.11" now), then:

git checkout -b dev/master-my-guile2
sh autogen.sh --noconfigure
mkdir -p build/
cd build/
../configure --enable-guile2
make -j5

I.e. simply compiling master with guilev2 and looking what happens: I
got a successful make.

Then trying to build the docs:
make LANGS='' doc
I.e. a single core compilation for better tracking problems, only english docs.

It crashed because of string-filter and string-delete changed the
order of arguments in guile2, I patched it and tried again.
Now it returns a segfault with:
/input/regression/markup-cyclic-reference.ly
Here I simply changed this regtest to put out null-markups, similiar
to what Antonio tried.

The complete diff:

address@hidden ~/lilypond-git (dev/master-my-guile2)$ git diff
diff --git a/input/regression/markup-cyclic-reference.ly
b/input/regression/markup-cyclic-reference.ly
index 82bfe06..993adfd 100644
--- a/input/regression/markup-cyclic-reference.ly
+++ b/input/regression/markup-cyclic-reference.ly
@@ -22,5 +22,5 @@ not crash LilyPond with an endless loop"
   (interpret-markup layout props (make-cycleI-markup m)))


-\markup { \cycle "a" }
-\markup { \cycleI "a" }
+\markup \null % { \cycle "a" }
+\markup \null % { \cycleI "a" }
diff --git a/scm/graphviz.scm b/scm/graphviz.scm
index 61fa5b1..43573d0 100644
--- a/scm/graphviz.scm
+++ b/scm/graphviz.scm
@@ -69,7 +69,7 @@
               es)
     (for-each (lambda (c)
                 (format out "subgraph cluster_~a {\nlabel=
\"~a\"\ncolor=blue\n"
-                        (string-filter (car c) char-alphabetic?)
+                        (string-filter char-alphabetic? (car c))
                         (car c))
                 (for-each (lambda (n) (format out "~a\n" n)) (cdr c))
                 (display "}\n" out))
diff --git a/scm/lily-library.scm b/scm/lily-library.scm
index 214c095..031c8be 100644
--- a/scm/lily-library.scm
+++ b/scm/lily-library.scm
@@ -779,8 +779,8 @@ as rectangular coordinates @code{(x-length . y-length)}."
 (define-public (remove-whitespace strg)
 "Remove characters satisfying @code{char-whitespace?} from string @var{strg}"
   (string-delete
-    strg
-    char-whitespace?))
+    char-whitespace?
+    strg))

 (define-public (string-encode-integer i)
   (cond
diff --git a/scm/lily.scm b/scm/lily.scm
index d4553ee..d3164e4 100644
--- a/scm/lily.scm
+++ b/scm/lily.scm
@@ -963,7 +963,7 @@ PIDs or the number of the process."
             (remove string-null?
                     (append-map
                      (lambda (f)
-                       (string-split (string-delete (ly:gulp-file f)
#\cr) #\nl))
+                       (string-split (string-delete #\cr
(ly:gulp-file f)) #\nl))
                      files))))
   (if (and (number? (ly:get-option 'job-count))
            (>= (length files) (ly:get-option 'job-count)))


I tried
make LANGS='' doc
again.

No more warnings or errors as far as I can tell, but it stucks at some point.
Now for more than 12h, but the single core stil works with 100%
I don't think it makes sense to try continuing for more hours! Or should I try?

Only doc-file so far created is collated-files.pdf (regtests). A
468-pages-pdf, sized 42.4 MB.
Though, I usually build and use the html-files. Thus I can't judge
whether this is normal or not.

Summary:
Ofcourse this is a very naive approach, using guile 2.0.11 not the
most current release.
Nevertheless I did not found any mentioning of string-delete and
string-filter being changed in previous discussions.
Apart from markup-cyclic-reference.ly the regtests compile (I did not
proof the output, though)

Any pointers how to continue?


Cheers,
  Harm

shortened terminal output attached

terminal-output.txt
Description: Text document