gnulib-tool test suite updates

[Collin Funk]

Hi Bruno,

On 3/30/24 4:22 PM, Bruno Haible wrote:
> Once the test suite entirely passes, will be the time to revisit this 
> 'cleaner'
> function. Your approach with the regex can be part of the solution.

Yes, that sounds like a good idea. It should be easier to validate
changes that way. The test cases are almost passing anyways. :)

I think I found some small issues in the test cases. Can you please
double check these for me? They should be easy to fix.

Here is what I see for the import tests at the moment:

$ env GNULIB_TOOL_IMPL=py ./test-all.sh 
PASS: test-hello-c-gnulib-1.sh
PASS: test-hello-c-gnulib-automakesubdir-1.sh
PASS: test-hello-c-gnulib-automakesubdir-withtests-1.sh
PASS: test-hello-c-gnulib-conddeps-1.sh
PASS: test-hello-c-gnulib-nonrecursive-1.sh
PASS: test-wget2-1.sh
./test-oath-toolkit-1.out tmp384427-out differ: byte 561, line 24
--- ./test-oath-toolkit-1.out   2024-03-30 14:11:44.586946254 -0700
+++ tmp384427-out       2024-03-30 17:14:43.330203522 -0700
@@ -21,6 +21,7 @@
   top/GNUmakefile
   top/maint.mk
 Updating file gl/m4/gnulib-common.m4 (backup in gl/m4/gnulib-common.m4~)
+Updating gl/Makefile.am (backup in gl/Makefile.am~)
 Finished.
 
 You may need to add #include directives for the following .h files.
FAIL: gnulib-tool's output has unexpected differences.
FAIL: test-oath-toolkit-1.sh
PASS: test-oath-toolkit-2.sh
PASS: test-oath-toolkit-3.sh
PASS: test-oath-toolkit-4.sh
PASS: test-oath-toolkit-5.sh
Files ./test-coreutils-1.result/gnulib-tests/gnulib.mk and 
tmp384554-result/gnulib-tests/gnulib.mk differ
Files ./test-coreutils-1.result/lib/gnulib.mk and 
tmp384554-result/lib/gnulib.mk differ
Files ./test-coreutils-1.result/m4/gnulib-cache.m4 and 
tmp384554-result/m4/gnulib-cache.m4 differ
Files ./test-coreutils-1.result/m4/gnulib-comp.m4 and 
tmp384554-result/m4/gnulib-comp.m4 differ
FAIL: gnulib-tool's result has unexpected differences.
FAIL: test-coreutils-1.sh
PASS: test-emacs-1.sh

I believe the 'test-oath-toolkit-1.sh' test just needs to be updated
since I see the same error when using gnulib-tool.sh.

The Coreutils one is from the --local-dir sorting I mentioned earlier.
I think I have a better idea on how to fix that now, so I'll submit it
soon.

Then when running the create-testdir tests:

$ env GNULIB_TOOL_IMPL=py ./test-all.sh 
PASS: test-create-testdir-1.sh
PASS: test-create-testdir-2.sh
PASS: test-create-testdir-3.sh
Files ./test-create-testdir-4.result/gllib/iconv_open-aix.h and 
tmp240808-result/gllib/iconv_open-aix.h differ
Files ./test-create-testdir-4.result/gllib/iconv_open-hpux.h and 
tmp240808-result/gllib/iconv_open-hpux.h differ
Files ./test-create-testdir-4.result/gllib/iconv_open-irix.h and 
tmp240808-result/gllib/iconv_open-irix.h differ
Files ./test-create-testdir-4.result/gllib/iconv_open-osf.h and 
tmp240808-result/gllib/iconv_open-osf.h differ
Files ./test-create-testdir-4.result/gllib/iconv_open-solaris.h and 
tmp240808-result/gllib/iconv_open-solaris.h differ
Files ./test-create-testdir-4.result/gllib/iconv_open-zos.h and 
tmp240808-result/gllib/iconv_open-zos.h differ
Files ./test-create-testdir-4.result/gllib/parse-datetime-gen.h and 
tmp240808-result/gllib/parse-datetime-gen.h differ
Files ./test-create-testdir-4.result/gllib/parse-datetime.c and 
tmp240808-result/gllib/parse-datetime.c differ
Files ./test-create-testdir-4.result/gllib/unicase/locale-languages.h and 
tmp240808-result/gllib/unicase/locale-languages.h differ
Files ./test-create-testdir-4.result/gllib/unicase/special-casing-table.h and 
tmp240808-result/gllib/unicase/special-casing-table.h differ
Files ./test-create-testdir-4.result/gllib/unictype/bidi_byname.h and 
tmp240808-result/gllib/unictype/bidi_byname.h differ
Files ./test-create-testdir-4.result/gllib/unictype/categ_byname.h and 
tmp240808-result/gllib/unictype/categ_byname.h differ
Files ./test-create-testdir-4.result/gllib/unictype/combiningclass_byname.h and 
tmp240808-result/gllib/unictype/combiningclass_byname.h differ
Files ./test-create-testdir-4.result/gllib/unictype/joininggroup_byname.h and 
tmp240808-result/gllib/unictype/joininggroup_byname.h differ
Files ./test-create-testdir-4.result/gllib/unictype/joiningtype_byname.h and 
tmp240808-result/gllib/unictype/joiningtype_byname.h differ
Files ./test-create-testdir-4.result/gllib/unictype/pr_byname.h and 
tmp240808-result/gllib/unictype/pr_byname.h differ
Files ./test-create-testdir-4.result/gllib/unictype/scripts_byname.h and 
tmp240808-result/gllib/unictype/scripts_byname.h differ
Files ./test-create-testdir-4.result/gllib/uninorm/composition-table.h and 
tmp240808-result/gllib/uninorm/composition-table.h differ
FAIL: gnulib-tool's result has unexpected differences.
FAIL: test-create-testdir-4.sh
PASS: test-create-megatestdir-1.sh
PASS: test-create-megatestdir-2.sh

These files are dependent on program versions. My gperf 3.1 package
had a wrong version number. I then built it from git master and saw
extra lines to silence warnings added in a few commits [1] [2]. I
checked out the 3.2 commit and, third times the charm, the iconv and
unicode files were fixed [3].

The 3.2 release commit didn't have a git tag and isn't on GNU's ftp
server. Not sure if you forgot or if this was intended, so I figured
I would let you know.

That leaves the parse-datetime files. The diff seems to explain
itself:

 /* Bison version string.  */
-#define YYBISON_VERSION "3.8.2"
+#define YYBISON_VERSION "3.8.2.46-9785-dirty"

I removed the one from my local path and everything is fine now. Maybe
we can remove these like the Autotools files? Or put the version
numbers in the README?

[1] 
https://git.savannah.gnu.org/gitweb/?p=gperf.git;a=commit;h=ce5106eb19a08ca9721641e8560969e73b7dd60e
[2] 
https://git.savannah.gnu.org/gitweb/?p=gperf.git;a=commit;h=ce5106eb19a08ca9721641e8560969e73b7dd60e
[3] 
https://git.savannah.gnu.org/gitweb/?p=gperf.git;a=commit;h=6d53a2dbe0ad692d0f36fd394b7e1962d9a81730
[4] https://ftp.gnu.org/gnu/gperf/

Collin