Bug#751822: Fix FTBFS on ppc64el (test-suite + python binding)

2014-08-14 Thread Andreas Metzler 
On 2014-06-17 Mauricio Faria de Oliveira  wrote:
[...]
> 2) python-libprelude-prefix.patch

>   This patch makes sure bindings/python/setup.py uses '/usr'
>   as prefix, since it can vary (occurence of '/usr/local' below).

>   If the prefix is different then /usr, a build failure occurs
>   because of a hardcoded '/usr/lib' in override_dh_auto_install:
>   

>   [...]
>debian/rules binary
>   dh binary --with=quilt,python2,autoreconf
>  [...]
>  debian/rules override_dh_auto_install
>  [...]
>   dh_auto_install; \
>   rm -rf debian/tmp/usr/lib/python*; \
>   find debian/tmp-python-libprelude/usr/lib -name "*.la" -delete; 
> \
>   mv debian/tmp-python-libprelude/usr/lib/* debian/tmp/usr/lib/
>   [...]
>   find: `debian/tmp-python-libprelude/usr/lib': No such file or directory
>   mv: cannot stat ‘debian/tmp-python-libprelude/usr/lib/*’: No such file 
> or
> directory
>   make[1]: *** [override_dh_auto_install] Error 1
>   [...]



Hello,

this patch looks strange. Can you show how to reproduce this issue?
This kind of bug cannot be ppc64el specific and I do not see it at
ix86.

cu Andreas

-- 
`What a good friend you are to him, Dr. Maturin. His other friends are
so grateful to you.'
`I sew his ears on from time to time, sure'


-- 
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org

Bug#751822: Fix FTBFS on ppc64el (test-suite + python binding)

2014-08-14 Thread Mauricio Faria de Oliveira 

Hi Andreas,

On 08/14/2014 02:24 PM, Andreas Metzler wrote:

this patch looks strange. Can you show how to reproduce this issue?
This kind of bug cannot be ppc64el specific and I do not see it at
ix86.


Yes, indeed strange.  I've realized it some weeks later, but missed
trying a bit more.

I'll try to reproduce/understand it again and ping back.

Thanks,

--
Mauricio Faria de Oliveira
IBM Linux Technology Center


--
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org

Bug#751822: Fix FTBFS on ppc64el (test-suite + python binding)

2014-08-15 Thread Mauricio Faria de Oliveira 

Hi Andreas,

On 08/14/2014 02:24 PM, Andreas Metzler wrote:
> this patch looks strange. Can you show how to reproduce this issue?

I wrote:
> I'll try to reproduce/understand it again and ping back.

Got it; it was due to building the package as a privileged user.
(thus passing setup.py's is_system_wide_install() test, due to
write-access on /usr/lib/python*/dist-packages)

The package builds fine on a normal/unprivileged sbuild build,
just with the test-suite patch.

I see you already made a uploaded w/ that patch for delayed/10.
Thanks, that will make it.

In case you'd prefer to put in a patch w/ DEP-3 headers and stuff,
I'm attaching one for your convenience.

Thanks again, and best regards.

--
Mauricio Faria de Oliveira
IBM Linux Technology Center
diff -Nru libprelude-1.0.0/debian/changelog libprelude-1.0.0/debian/changelog
--- libprelude-1.0.0/debian/changelog   2014-08-05 17:29:12.0 +
+++ libprelude-1.0.0/debian/changelog   2014-08-15 16:29:33.0 +
@@ -1,3 +1,9 @@
+libprelude (1.0.0-11.3ppc64el1) UNRELEASED; urgency=medium
+
+  * Fix libmissing/ test-suite on ppc64el (ppc64el-fix-gnulib-tests.patch) 
(Closes: #751822)
+
+ -- Mauricio Faria de Oliveira   Thu, 14 Aug 2014 
18:06:04 -0300
+
 libprelude (1.0.0-11.3) unstable; urgency=low
 
   * Non-maintainer upload.
diff -Nru libprelude-1.0.0/debian/patches/ppc64el-fix-gnulib-tests.patch 
libprelude-1.0.0/debian/patches/ppc64el-fix-gnulib-tests.patch
--- libprelude-1.0.0/debian/patches/ppc64el-fix-gnulib-tests.patch  
1970-01-01 00:00:00.0 +
+++ libprelude-1.0.0/debian/patches/ppc64el-fix-gnulib-tests.patch  
2014-08-15 16:28:09.0 +
@@ -0,0 +1,55 @@
+Subject: [PATCH] isfinite, isinf, isnan tests: fix for little-endian PowerPC
+Origin: backport, 
http://git.savannah.gnu.org/gitweb/?p=gnulib.git;a=commit;h=e962c07c0c8bb18f9a5969cbf1ddb01d0fe4f55c
+Forwarded: https://www.prelude-ids.org/issues/556
+Author: Mauricio Faria de Oliveira 
+Last-Update: 2014-08-14
+
+This patch is a trivial backport (subset) of the upstream patch; it just 
removes
+the files that don't exist in libprelude (not required).
+
+ curl 
'http://git.savannah.gnu.org/gitweb/?p=gnulib.git;a=patch;h=e962c07c0c8bb18f9a5969cbf1ddb01d0fe4f55c'
 \
+  | filterdiff \
+  --strip=1 \
+  --addoldprefix='a/libmissing/' \
+  --addnewprefix='b/libmissing/' \
+  -i '*/tests/test-isnanl.h' \
+  -i '*/tests/test-signbit.c' \
+  > ppc64el-fix-gnulib-tests.patch
+
+--- a/libmissing/tests/test-isnanl.h
 b/libmissing/tests/test-isnanl.h
+@@ -51,6 +51,15 @@ main ()
+   /* A bit pattern that is different from a Quiet NaN.  With a bit of luck,
+  it's a Signalling NaN.  */
+   {
++#if defined __powerpc__ && LDBL_MANT_DIG == 106
++/* This is PowerPC "double double", a pair of two doubles.  Inf and Nan 
are
++   represented as the corresponding 64-bit IEEE values in the first 
double;
++   the second is ignored.  Manipulate only the first double.  */
++#undef NWORDS
++#define NWORDS \
++  ((sizeof (double) + sizeof (unsigned int) - 1) / sizeof (unsigned int))
++#endif
++
+ memory_long_double m;
+ m.value = NaNl ();
+ # if LDBL_EXPBIT0_BIT > 0
+--- a/libmissing/tests/test-signbit.c
 b/libmissing/tests/test-signbit.c
+@@ -151,6 +151,16 @@ test_signbitl ()
+ #define NWORDS \
+   ((sizeof (long double) + sizeof (unsigned int) - 1) / sizeof (unsigned 
int))
+ typedef union { long double value; unsigned int word[NWORDS]; } 
memory_long_double;
++
++#if defined __powerpc__ && LDBL_MANT_DIG == 106
++/* This is PowerPC "double double", a pair of two doubles.  Inf and Nan 
are
++   represented as the corresponding 64-bit IEEE values in the first 
double;
++   the second is ignored.  Manipulate only the first double.  */
++#undef NWORDS
++#define NWORDS \
++  ((sizeof (double) + sizeof (unsigned int) - 1) / sizeof (unsigned int))
++#endif
++
+ memory_long_double m;
+ m.value = zerol / zerol;
+ # if LDBL_EXPBIT0_BIT > 0
diff -Nru libprelude-1.0.0/debian/patches/series 
libprelude-1.0.0/debian/patches/series
--- libprelude-1.0.0/debian/patches/series  2014-05-25 13:58:53.0 
+
+++ libprelude-1.0.0/debian/patches/series  2014-08-15 16:28:09.0 
+
@@ -5,3 +5,4 @@
 request_REUSEADDR_before_bind.patch
 test-poll_add_sleep_after_write.patch
 keep-atfork-symbol.patch
+ppc64el-fix-gnulib-tests.patch

Bug#751822: Fix FTBFS on ppc64el (test-suite + python binding)

2014-06-16 Thread Mauricio Faria de Oliveira 

Package: src:libprelude
Version: 1.0.0-1.0.0-11.2
Severity: normal
Tags: patch
User: debian-powe...@lists.debian.org
Usertags: ppc64el

Hi, 

These 2 patches address build failures of libprelude on ppc64el.
May you please consider incorporating them into an upload?

Thank you!


1) ppc64el-isnan-tests-fix-for-little-endian-PowerPC.patch

This patch is upstream at gnulib as of May 30th, 2014
(commit e962c07c0c8bb18f9a5969cbf1ddb01d0fe4f55c).

It picks the subset applicable to files carried by libprelude.

It fixes this test-suite error:

PASS: test-isnand-nolibm
PASS: test-isnanf-nolibm
		../../test-driver: line 107: 20059 Aborted "$@" > 
$log_file 2>&1

FAIL: test-isnanl-nolibm



2) python-libprelude-prefix.patch

This patch makes sure bindings/python/setup.py uses '/usr'
as prefix, since it can vary (occurence of '/usr/local' below).

If the prefix is different then /usr, a build failure occurs
because of a hardcoded '/usr/lib' in override_dh_auto_install:


[...]
 debian/rules binary
dh binary --with=quilt,python2,autoreconf
   [...]
   debian/rules override_dh_auto_install
   [...]
dh_auto_install; \
rm -rf debian/tmp/usr/lib/python*; \
find debian/tmp-python-libprelude/usr/lib -name "*.la" -delete; 
\
mv debian/tmp-python-libprelude/usr/lib/* debian/tmp/usr/lib/
[...]
find: `debian/tmp-python-libprelude/usr/lib': No such file or directory
	mv: cannot stat ‘debian/tmp-python-libprelude/usr/lib/*’: No such file 
or directory

make[1]: *** [override_dh_auto_install] Error 1
[...]



--
Mauricio Faria de Oliveira
IBM Linux Technology Center
Index: libprelude-1.0.0/bindings/python/Makefile.am
===
--- libprelude-1.0.0.orig/bindings/python/Makefile.am
+++ libprelude-1.0.0/bindings/python/Makefile.am
@@ -9,7 +9,7 @@
 	CC="$(CXX)" $(PYTHON) setup.py build
 
 install-exec-hook:
-	$(PYTHON) setup.py install `test -n "$(DESTDIR)" && echo --root $(DESTDIR)`
+	$(PYTHON) setup.py install `test -n "$(DESTDIR)" && echo --root $(DESTDIR)` --prefix /usr
 
 uninstall-hook:
 	$(PYTHON) setup.py uninstall `test -n "$(DESTDIR)" && echo --root $(DESTDIR)`
curl 'http://git.savannah.gnu.org/gitweb/?p=gnulib.git;a=patch;h=e962c07c0c8bb18f9a5969cbf1ddb01d0fe4f55c' \
  | filterdiff \
  --strip=1 \
  --addoldprefix='a/libmissing/' \
  --addnewprefix='b/libmissing/' \
  -i '*/tests/test-isnanl.h' \
  -i '*/tests/test-signbit.c' \
  > 'ppc64el-isnan-tests-fix-for-little-endian-PowerPC.patch'

--- a/libmissing/tests/test-isnanl.h
+++ b/libmissing/tests/test-isnanl.h
@@ -51,6 +51,15 @@ main ()
   /* A bit pattern that is different from a Quiet NaN.  With a bit of luck,
  it's a Signalling NaN.  */
   {
+#if defined __powerpc__ && LDBL_MANT_DIG == 106
+/* This is PowerPC "double double", a pair of two doubles.  Inf and Nan are
+   represented as the corresponding 64-bit IEEE values in the first double;
+   the second is ignored.  Manipulate only the first double.  */
+#undef NWORDS
+#define NWORDS \
+  ((sizeof (double) + sizeof (unsigned int) - 1) / sizeof (unsigned int))
+#endif
+
 memory_long_double m;
 m.value = NaNl ();
 # if LDBL_EXPBIT0_BIT > 0
--- a/libmissing/tests/test-signbit.c
+++ b/libmissing/tests/test-signbit.c
@@ -151,6 +151,16 @@ test_signbitl ()
 #define NWORDS \
   ((sizeof (long double) + sizeof (unsigned int) - 1) / sizeof (unsigned int))
 typedef union { long double value; unsigned int word[NWORDS]; } memory_long_double;
+
+#if defined __powerpc__ && LDBL_MANT_DIG == 106
+/* This is PowerPC "double double", a pair of two doubles.  Inf and Nan are
+   represented as the corresponding 64-bit IEEE values in the first double;
+   the second is ignored.  Manipulate only the first double.  */
+#undef NWORDS
+#define NWORDS \
+  ((sizeof (double) + sizeof (unsigned int) - 1) / sizeof (unsigned int))
+#endif
+
 memory_long_double m;
 m.value = zerol / zerol;
 # if LDBL_EXPBIT0_BIT > 0