bug#10681: GNU Guile 2.0.5 released

bug#10681: GNU Guile 2.0.5 released

[Hans Aberg]

On 30 Jan 2012, at 23:02, Ludovic Courtès wrote:

> We are pleased to announce GNU Guile release 2.0.5.

The 'make check' gives this error, on OS X 10.7.2 using the compiler i686-apple-darwin11-llvm-gcc-4.2 (GCC) 4.2.1 supplied by Xcode 4.2.1:

  PASS: test-asmobs
bad return from expression `(f-sum -1 2000 -30000 40000000000)': expected 39999971999; got 39999972255
/bin/sh: line 1: 96687 Bus error: 10           srcdir="." builddir="." CHARSETALIASDIR="/usr/local/src/guile/guile-2.0.5/lib" GUILE_AUTO_COMPILE=0 "../../meta/uninstalled-env" ${dir}$tst
...
==================================
1 of 28 tests failed
Please report to bug-guile@gnu.org
==================================

Done.

Hans

bug#10681: GNU Guile 2.0.5 released

[Andy Wingo]

https://savannah.gnu.org/bugs/?30122

-- 
http://wingolog.org/

bug#10681: GNU Guile 2.0.5 released

[Hans Aberg]

On 31 Jan 2012, at 15:40, Andy Wingo wrote:

> https://savannah.gnu.org/bugs/?30122


I used /usr/bin/gcc -> llvm-gcc-4.2, which is different from clang.

There is also this one
  http://lists.gnu.org/archive/html/bug-guile/2011-11/msg00026.html

It asks for this output:
$ grep scm_t_int8 libguile/scmconfig.h
typedef int8_t scm_t_int8;

It is the same with gcc-4.7.0.

Hans

bug#10681: GNU Guile 2.0.5 released

[Ludovic Courtès]

Hi Hans,

Hans Aberg <haberg-1@telia.com> skribis:

> The 'make check' gives this error, on OS X 10.7.2 using the compiler i686-apple-darwin11-llvm-gcc-4.2 (GCC) 4.2.1 supplied by Xcode 4.2.1:

What is this?  Apple’s GCC?  DragonEgg?

FWIW, this problem doesn’t show up on
<http://hydra.nixos.org/jobset/gnu/guile-2-0/>, which uses Apple’s GCC
4.2.1 on x86_64-apple-darwin10.2.0.

Thanks,
Ludo’.

bug#10681: GNU Guile 2.0.5 released

[Hans Aberg]

On 31 Jan 2012, at 16:18, Ludovic Courtès wrote:

>> The 'make check' gives this error, on OS X 10.7.2 using the compiler i686-apple-darwin11-llvm-gcc-4.2 (GCC) 4.2.1 supplied by Xcode 4.2.1:
> 
> What is this?  Apple’s GCC?  DragonEgg?

On OS X 10.7.2, Xcode 4.2 installs two system compilers
  /usr/bin/clang
  /usr/bin/gcc -> llvm-gcc-4.2
  /usr/bin/cc -> llvm-gcc-4.2
It is the latter, because that is one gets hold of without 'export CC=...'.

> FWIW, this problem doesn’t show up on
> <http://hydra.nixos.org/jobset/gnu/guile-2-0/>, which uses Apple’s GCC
> 4.2.1 on x86_64-apple-darwin10.2.0.

I get another error with gcc (GCC) 4.7.0 (from SVN, installed in /usr/local/bin/gcc):

PASS: test-asmobs
/bin/sh: line 1: 33654 Bus error: 10           srcdir="." builddir="." CHARSETALIASDIR="/usr/local/src/guile/gcc-4.7/guile-2.0.5/lib" GUILE_AUTO_COMPILE=0 "../../meta/uninstalled-env" ${dir}$tst
FAIL: test-ffi
PASS: test-list
...
==================================
1 of 28 tests failed
Please report to bug-guile@gnu.org
==================================

So the f-sum error seems to have to do with llvm-gcc (and possibly clang).

In both cases, the failure is
  FAIL: test-ffi

Hans

bug#10681: GNU Guile 2.0.5 released

[Mark H Weaver]

Hans Aberg <haberg-1@telia.com> writes:

> The 'make check' gives this error, on OS X 10.7.2 using the compiler
> i686-apple-darwin11-llvm-gcc-4.2 (GCC) 4.2.1 supplied by Xcode 4.2.1:
>
>   PASS: test-asmobs
> bad return from expression `(f-sum -1 2000 -30000 40000000000)': expected 39999971999; got 39999972255

I may be stating the obvious here, but the -1, which is declared to be
of type 'scm_t_int8' in the C function being called, is apparently being
interpreted as 255.  This suggests that whatever is handling its
promotion to a full int is failing to extend its sign bit.  I'm guessing
that this is libffi's job.

Hans: can you please verify that your libffi's 'make check' passes all
tests on your platform?  If it passes, and if this ends up being a
different build of 'libffi' than you were previously using, it would be
helpful if you could install the newly-built 'libffi', then do a fresh
rebuild of Guile 2.0.5 and see if that fixes the problem.

     Thanks,
       Mark

bug#10681: GNU Guile 2.0.5 released

[Hans Aberg]

On 31 Jan 2012, at 19:04, Mark H Weaver wrote:

>> The 'make check' gives this error, on OS X 10.7.2 using the compiler
>> i686-apple-darwin11-llvm-gcc-4.2 (GCC) 4.2.1 supplied by Xcode 4.2.1:
>> 
>>  PASS: test-asmobs
>> bad return from expression `(f-sum -1 2000 -30000 40000000000)': expected 39999971999; got 39999972255
> 
> I may be stating the obvious here, but the -1, which is declared to be
> of type 'scm_t_int8' in the C function being called, is apparently being
> interpreted as 255.  This suggests that whatever is handling its
> promotion to a full int is failing to extend its sign bit.  I'm guessing
> that this is libffi's job.

It is broken (libffi from GIT, the only that works on OX 10.7):

$ make check
Making check in include
make[1]: Nothing to be done for `check'.
Making check in testsuite
make  check-DEJAGNU
srcdir=`CDPATH="${ZSH_VERSION+.}:" && cd ../../libffi/testsuite && pwd`; export srcdir; \
	EXPECT=`if [ -f ../../expect/expect ] ; then echo ../../expect/expect ; else echo expect ; fi`; export EXPECT; \
	runtest=`if [ -f ../../libffi/../dejagnu/runtest ] ; then echo ../../libffi/../dejagnu/runtest ; else echo runtest; fi`; \
	if /bin/sh -c "$runtest --version" > /dev/null 2>&1; then \
	  exit_status=0; l='libffi'; for tool in $l; do \
	    if $runtest  --tool $tool --srcdir $srcdir ; \
	    then :; else exit_status=1; fi; \
	  done; \
	else echo "WARNING: could not find \`runtest'" 1>&2; :;\
	fi; \
	exit $exit_status
WARNING: could not find `runtest'
Making check in man
make[1]: Nothing to be done for `check'.
make[1]: Nothing to be done for `check-am'

> Hans: can you please verify that your libffi's 'make check' passes all
> tests on your platform?  If it passes, and if this ends up being a
> different build of 'libffi' than you were previously using, it would be
> helpful if you could install the newly-built 'libffi', then do a fresh
> rebuild of Guile 2.0.5 and see if that fixes the problem.

I have installed the newly built libffi, but rebuilding guile takes a lot of time.

Hans

bug#10681: GNU Guile 2.0.5 released

[Mark H Weaver]

Hans Aberg <haberg-1@telia.com> writes:

> On 31 Jan 2012, at 19:04, Mark H Weaver wrote:
>
>>> The 'make check' gives this error, on OS X 10.7.2 using the compiler
>>> i686-apple-darwin11-llvm-gcc-4.2 (GCC) 4.2.1 supplied by Xcode 4.2.1:
>>> 
>>>  PASS: test-asmobs
>>> bad return from expression `(f-sum -1 2000 -30000 40000000000)': expected 39999971999; got 39999972255
>> 
>> I may be stating the obvious here, but the -1, which is declared to be
>> of type 'scm_t_int8' in the C function being called, is apparently being
>> interpreted as 255.  This suggests that whatever is handling its
>> promotion to a full int is failing to extend its sign bit.  I'm guessing
>> that this is libffi's job.
>
> It is broken (libffi from GIT, the only that works on OX 10.7):

You need to install DejaGnu in order to run libffi's test suite.
<http://www.gnu.org/software/dejagnu/>

   Thanks,
     Mark

bug#10681: GNU Guile 2.0.5 released

[Hans Aberg]

On 31 Jan 2012, at 20:35, Mark H Weaver wrote:

>>>> The 'make check' gives this error, on OS X 10.7.2 using the compiler
>>>> i686-apple-darwin11-llvm-gcc-4.2 (GCC) 4.2.1 supplied by Xcode 4.2.1:
>>>> 
>>>> PASS: test-asmobs
>>>> bad return from expression `(f-sum -1 2000 -30000 40000000000)': expected 39999971999; got 39999972255
>>> 
>>> I may be stating the obvious here, but the -1, which is declared to be
>>> of type 'scm_t_int8' in the C function being called, is apparently being
>>> interpreted as 255.  This suggests that whatever is handling its
>>> promotion to a full int is failing to extend its sign bit.  I'm guessing
>>> that this is libffi's job.
>> 
>> It is broken (libffi from GIT, the only that works on OX 10.7):
> 
> You need to install DejaGnu in order to run libffi's test suite.
> <http://www.gnu.org/software/dejagnu/>

OK. I am rebuilding Guile. Stay tuned.

Hans

bug#10681: GNU Guile 2.0.5 released

[Hans Aberg]

On 31 Jan 2012, at 20:35, Mark H Weaver wrote:

>>>> The 'make check' gives this error, on OS X 10.7.2 using the compiler
>>>> i686-apple-darwin11-llvm-gcc-4.2 (GCC) 4.2.1 supplied by Xcode 4.2.1:
>>>> 
>>>> PASS: test-asmobs
>>>> bad return from expression `(f-sum -1 2000 -30000 40000000000)': expected 39999971999; got 39999972255
>>> 
>>> I may be stating the obvious here, but the -1, which is declared to be
>>> of type 'scm_t_int8' in the C function being called, is apparently being
>>> interpreted as 255.  This suggests that whatever is handling its
>>> promotion to a full int is failing to extend its sign bit.  I'm guessing
>>> that this is libffi's job.
>> 
>> It is broken (libffi from GIT, the only that works on OX 10.7):
> 
> You need to install DejaGnu in order to run libffi's test suite.
> <http://www.gnu.org/software/dejagnu/>

It seem they passed (not giving any details).

Hans


Test Run on Tue Jan 31 20:51:33 2012
Native configuration is x86_64-apple-darwin11.2.0

		=== libffi tests ===

Schedule of variations:
    unix

Running target unix
Using /usr/local/share/dejagnu/baseboards/unix.exp as board description file for target.
Using /usr/local/share/dejagnu/config/unix.exp as generic interface file for target.
Using /usr/local/src/libffi/git/libffi/testsuite/config/default.exp as tool-and-target-specific interface file.
Running /usr/local/src/libffi/git/libffi/testsuite/libffi.call/call.exp ...
Running /usr/local/src/libffi/git/libffi/testsuite/libffi.special/special.exp ...

		=== libffi Summary ===

# of expected passes		1659
# of unsupported tests		15
Making check in man
make[1]: Nothing to be done for `check'.
make[1]: Nothing to be done for `check-am'.

bug#10681: GNU Guile 2.0.5 released

[Hans Aberg]

On 31 Jan 2012, at 20:35, Mark H Weaver wrote:

>>>> The 'make check' gives this error, on OS X 10.7.2 using the compiler
>>>> i686-apple-darwin11-llvm-gcc-4.2 (GCC) 4.2.1 supplied by Xcode 4.2.1:
>>>> 
>>>> PASS: test-asmobs
>>>> bad return from expression `(f-sum -1 2000 -30000 40000000000)': expected 39999971999; got 39999972255
>>> 
>>> I may be stating the obvious here, but the -1, which is declared to be
>>> of type 'scm_t_int8' in the C function being called, is apparently being
>>> interpreted as 255.  This suggests that whatever is handling its
>>> promotion to a full int is failing to extend its sign bit.  I'm guessing
>>> that this is libffi's job.
>> 
>> It is broken (libffi from GIT, the only that works on OX 10.7):
> 
> You need to install DejaGnu in order to run libffi's test suite.
> <http://www.gnu.org/software/dejagnu/>

With gcc-4.7.0 (from SVN), the test-ffi test now passes (libffi from GIT), but I get three other failures.

The compiler that is normally used on the system, is llvm-gcc-4.2, and its compile is still running.

Hans


Running bytevectors.test
FAIL: bytevectors.test: 2.3 Operations on Bytes and Octets: bytevector-sint-ref [small] (eval)
FAIL: bytevectors.test: 2.3 Operations on Bytes and Octets: bytevector-sint-ref [small] (compile)
...
Running gc.test
FAIL: gc.test: gc: Unused modules are removed
...

Totals for this test run:
passes:                 34886
failures:               3
unexpected passes:      0
expected failures:      30
unresolved test cases:  29
untested test cases:    1
unsupported test cases: 9
errors:                 0

FAIL: check-guile
==================================
1 of 1 test failed
Please report to bug-guile@gnu.org
==================================

bug#10681: GNU Guile 2.0.5 released

[Mark H Weaver]

[-- Attachment #1: Type: text/plain, Size: 1180 bytes --]

Hans Aberg <haberg-1@telia.com> writes:
> With gcc-4.7.0 (from SVN), the test-ffi test now passes (libffi from
> GIT)

Excellent!  I guess that this was a libffi bug.

> but I get three other failures.
>
> The compiler that is normally used on the system, is llvm-gcc-4.2, and
> its compile is still running.

Please let us know the results of 'make check' when compiling with
llvm-gcc-4.2.  I'm especially curious to hear whether the bytevector
tests fail with that compiler as well.

> Running bytevectors.test
> FAIL: bytevectors.test: 2.3 Operations on Bytes and Octets: bytevector-sint-ref [small] (eval)
> FAIL: bytevectors.test: 2.3 Operations on Bytes and Octets: bytevector-sint-ref [small] (compile)

In the directory where you built using GCC-4.7.0 (SVN), can you please
apply the following 'patch for bytevectors.test' and then, from the
guile-2.0.5 directory, run:

  ./check-guile bytevectors.test

and show us the output?

This is a shot in the dark, but I've also attached a patch that _might_
fix the bytevector problem.  After applying it, it should be sufficient
to simply run "make" again, and it shouldn't take long.  Let us know!

     Thanks,
       Mark



[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: Patch for bytevectors.test (to help diagnose) --]
[-- Type: text/x-patch, Size: 1028 bytes --]

diff --git a/test-suite/tests/bytevectors.test b/test-suite/tests/bytevectors.test
index 3007434..b652935 100644
--- a/test-suite/tests/bytevectors.test
+++ b/test-suite/tests/bytevectors.test
@@ -114,10 +114,14 @@
               #xfffffffffffffffffffffffffffffffd)))
 
   (pass-if "bytevector-sint-ref [small]"
-    (let ((b (u8-list->bytevector '(#xff #xf0 #xff))))
-      (equal? (bytevector-sint-ref b 0 (endianness big) 2)
-              (bytevector-sint-ref b 1 (endianness little) 2)
-              -16)))
+    (let* ((b (u8-list->bytevector '(#xff #xf0 #xff)))
+           (be-result (bytevector-sint-ref b 0 (endianness big) 2))
+           (le-result (bytevector-sint-ref b 1 (endianness little) 2)))
+      (or (equal? be-result le-result -16)
+          (begin (format (current-error-port)
+                         "bytevector-sint-ref [small] failure: ~S ~S~%"
+                         be-result le-result)
+                 #f))))
 
   (pass-if "bytevector-sint-ref [large]"
     (let ((b (make-bytevector 50)))

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #3: A shot in the dark (possible fix) --]
[-- Type: text/x-patch, Size: 1948 bytes --]

diff --git a/libguile/bytevectors.c b/libguile/bytevectors.c
index fff5355..8574a36 100644
--- a/libguile/bytevectors.c
+++ b/libguile/bytevectors.c
@@ -103,7 +103,7 @@
                                                                 \
     memcpy (&c_result, &c_bv[c_index], (_len) / 8);             \
     if (!scm_is_eq (endianness, scm_i_native_endianness))       \
-      c_result = INT_SWAP (_len) (c_result);                    \
+      c_result = (INT_TYPE (_len, _sign)) INT_SWAP (_len) ((INT_TYPE (_len, unsigned)) c_result); \
                                                                 \
     result = SCM_I_MAKINUM (c_result);                          \
   }                                                             \
@@ -143,7 +143,7 @@
 								\
     c_value_short = (INT_TYPE (_len, _sign)) c_value;		\
     if (!scm_is_eq (endianness, scm_i_native_endianness))       \
-      c_value_short = INT_SWAP (_len) (c_value_short);		\
+      c_value_short = (INT_TYPE (_len, _sign)) INT_SWAP (_len) ((INT_TYPE (_len, unsigned)) c_value_short);  \
 								\
     memcpy (&c_bv[c_index], &c_value_short, (_len) / 8);	\
   }								\
@@ -918,7 +918,7 @@ bytevector_large_set (char *c_bv, size_t c_size, int signed_p,
 	    INT_TYPE (16, _sign)  c_value16;				\
 	    memcpy (&c_value16, c_bv, 2);				\
 	    if (swap)							\
-	      value = (INT_TYPE (16, _sign)) bswap_16 (c_value16);	\
+	      value = (INT_TYPE (16, _sign)) bswap_16 ((scm_t_uint16) c_value16); \
 	    else							\
 	      value = c_value16;					\
 	  }								\
@@ -981,7 +981,7 @@ bytevector_unsigned_ref (const char *c_bv, size_t c_size, SCM endianness)
 	      swap = !scm_is_eq (endianness, scm_i_native_endianness);	\
 									\
 	      if (swap)							\
-		c_value16 = (INT_TYPE (16, _sign)) bswap_16 (c_value);	\
+		c_value16 = (INT_TYPE (16, _sign)) bswap_16 ((scm_t_uint16) c_value); \
 	      else							\
 		c_value16 = c_value;					\
 									\

bug#10681: GNU Guile 2.0.5 released

[Hans Aberg]

On 1 Feb 2012, at 02:42, Mark H Weaver wrote:

> Hans Aberg <haberg-1@telia.com> writes:
>> With gcc-4.7.0 (from SVN), the test-ffi test now passes (libffi from
>> GIT)
> 
> Excellent!  I guess that this was a libffi bug.

No, I think it is with llvm-gcc, in view of that it remained in that compile (as described in another letter).

>> but I get three other failures.
>> 
>> The compiler that is normally used on the system, is llvm-gcc-4.2, and
>> its compile is still running.
> 
> Please let us know the results of 'make check' when compiling with
> llvm-gcc-4.2.  I'm especially curious to hear whether the bytevector
> tests fail with that compiler as well.

There is no such test with the llvm-gcc compiler, strangely enough: it just produces a few tests. It does not show the header that is shown for gcc-4.7 (below), like this

Hans


From gcc-4.7 compile 'make check':

Totals for this test run:
passes:                 34886
failures:               3
unexpected passes:      0
expected failures:      30
unresolved test cases:  29
untested test cases:    1
unsupported test cases: 9
errors:                 0

bug#10681: GNU Guile 2.0.5 released

[Hans Aberg]

On 1 Feb 2012, at 02:42, Mark H Weaver wrote:

>> Running bytevectors.test
>> FAIL: bytevectors.test: 2.3 Operations on Bytes and Octets: bytevector-sint-ref [small] (eval)
>> FAIL: bytevectors.test: 2.3 Operations on Bytes and Octets: bytevector-sint-ref [small] (compile)
> 
> In the directory where you built using GCC-4.7.0 (SVN), can you please
> apply the following 'patch for bytevectors.test' …

This assumes that one builds from within the source directory. 

> and then, from the
> guile-2.0.5 directory, run:
> 
>  ./check-guile bytevectors.test
> 
> and show us the output?

It did not help (output below). 

Hans


$ ./check-guile bytevectors.test
Testing /usr/local/src/guile/gcc-4.7/guile-2.0.5-build/meta/guile ... bytevectors.test
with GUILE_LOAD_PATH=/usr/local/src/guile/gcc-4.7/guile-2.0.5/test-suite
Running bytevectors.test
bytevector-sint-ref [small] failure: -16 4294967280
FAIL: bytevectors.test: 2.3 Operations on Bytes and Octets: bytevector-sint-ref [small] (eval)
bytevector-sint-ref [small] failure: -16 4294967280
FAIL: bytevectors.test: 2.3 Operations on Bytes and Octets: bytevector-sint-ref [small] (compile)

Totals for this test run:
passes:                 132
failures:               2
unexpected passes:      0
expected failures:      0
unresolved test cases:  0
untested test cases:    0
unsupported test cases: 0
errors:                 0

bug#10681: GNU Guile 2.0.5 released

[Hans Aberg]

On 31 Jan 2012, at 20:35, Mark H Weaver wrote:

>>>> The 'make check' gives this error, on OS X 10.7.2 using the compiler
>>>> i686-apple-darwin11-llvm-gcc-4.2 (GCC) 4.2.1 supplied by Xcode 4.2.1:
>>>> 
>>>> PASS: test-asmobs
>>>> bad return from expression `(f-sum -1 2000 -30000 40000000000)': expected 39999971999; got 39999972255
>>> 
>>> I may be stating the obvious here, but the -1, which is declared to be
>>> of type 'scm_t_int8' in the C function being called, is apparently being
>>> interpreted as 255.  This suggests that whatever is handling its
>>> promotion to a full int is failing to extend its sign bit.  I'm guessing
>>> that this is libffi's job.
>> 
>> It is broken (libffi from GIT, the only that works on OX 10.7):
> 
> You need to install DejaGnu in order to run libffi's test suite.
> <http://www.gnu.org/software/dejagnu/>

After doing this, the same failure with the LLVM-GCC compiler:
  /usr/bin/cc -> llvm-gcc-4.2
  /usr/bin/gcc -> llvm-gcc-4.2
  i686-apple-darwin11-llvm-gcc-4.2 (GCC) 4.2.1

This is the compiler that one will use on OS X 10.7 if one installs Xcode 4.2.1, and is not setting the compiler explicitly (or overriding by another install).

Hans


PASS: test-asmobs
bad return from expression `(f-sum -1 2000 -30000 40000000000)': expected 39999971999; got 39999972255
FAIL: test-ffi
...
==================================
1 of 28 tests failed
Please report to bug-guile@gnu.org
==================================

bug#10681: GNU Guile 2.0.5 released

[Mark H Weaver]

Hans Aberg <haberg-1@telia.com> writes:
> After doing this, the same failure with the LLVM-GCC compiler:
>   /usr/bin/cc -> llvm-gcc-4.2
>   /usr/bin/gcc -> llvm-gcc-4.2
>   i686-apple-darwin11-llvm-gcc-4.2 (GCC) 4.2.1
>
> This is the compiler that one will use on OS X 10.7 if one installs
> Xcode 4.2.1, and is not setting the compiler explicitly (or overriding
> by another install).
>
> Hans
>
>
> PASS: test-asmobs
> bad return from expression `(f-sum -1 2000 -30000 40000000000)': expected 39999971999; got 39999972255
> FAIL: test-ffi

Are you sure this guile was linked against your newly-built 'libffi'?
Your previously reported results (using GCC 4.7.0 from SVN) seemed to
suggest a bug in an earlier version of 'libffi' that has since been
fixed in their development tree.

     Mark

bug#10681: GNU Guile 2.0.5 released

[Hans Aberg]

On 1 Feb 2012, at 02:49, Mark H Weaver wrote:

>> After doing this, the same failure with the LLVM-GCC compiler:
>>  /usr/bin/cc -> llvm-gcc-4.2
>>  /usr/bin/gcc -> llvm-gcc-4.2
>>  i686-apple-darwin11-llvm-gcc-4.2 (GCC) 4.2.1
>> 
>> This is the compiler that one will use on OS X 10.7 if one installs
>> Xcode 4.2.1, and is not setting the compiler explicitly (or overriding
>> by another install).
>> 
>> Hans
>> 
>> 
>> PASS: test-asmobs
>> bad return from expression `(f-sum -1 2000 -30000 40000000000)': expected 39999971999; got 39999972255
>> FAIL: test-ffi
> 
> Are you sure this guile was linked against your newly-built 'libffi'?

Yes, I compiled it with llvm-gcc, and installed it. All new compiles were made out of the source directory.

> Your previously reported results (using GCC 4.7.0 from SVN) seemed to
> suggest a bug in an earlier version of 'libffi' that has since been
> fixed in their development tree.

It suggests that problem is with llvm-gcc (an clang), I think. With gcc-4.7 there is no libffi failure.

Hans

bug#10681: GNU Guile 2.0.5 released

[Andy Wingo]

On Wed 01 Feb 2012 10:18, Hans Aberg <haberg-1@telia.com> writes:

> It suggests that problem is with llvm-gcc (an clang), I think. With
> gcc-4.7 there is no libffi failure.

Is it correct to say that you experience this issue if libffi is
compiled with llvm-gcc / clang, but do not experience this issue if
libffi is compiled with gcc-4.7 ?

Trying to summarize; it has been hard to follow :)

Andy
-- 
http://wingolog.org/

bug#10681: GNU Guile 2.0.5 released

[Hans Aberg]

On 1 Feb 2012, at 12:50, Andy Wingo wrote:

>> It suggests that problem is with llvm-gcc (an clang), I think. With
>> gcc-4.7 there is no libffi failure.
> 
> Is it correct to say that you experience this issue if libffi is
> compiled with llvm-gcc / clang, …

Yes, and also guile-2.0.5 (see below for more info).

> ...but do not experience this issue if
> libffi is compiled with gcc-4.7 ?

I have not tried that one. There is not issue with libffi from latest GIT compiled with llvm-gcc-4.2, and guile-2.0.5 compiled with SVN gcc-4.7.

> Trying to summarize; it has been hard to follow :)

Indeed, as there are three compilers :-):
  /usr/bin/clang
  /usr/bin/cc -> llvm-gcc-4.2
  /usr/bin/gcc -> llvm-gcc-4.2
These are provided by Xcode 4.2.1. So if one does not set CC, one will use llvm-gcc-4.2.

This compiler, llvm-gcc-4.2, is also what I use for system installation, as the SVN gcc-4.7 is experimental. Also gcc-4.7 was compiled using llvm-gcc-4.2; it did not compile with gcc-4.6.2.

So it seems safest to stick to llvm-gcc-4.2, as that is what package developers mostly will check against.

Hans

bug#10681: GNU Guile 2.0.5 released

[Andy Wingo]

On Wed 01 Feb 2012 14:36, Hans Aberg <haberg-1@telia.com> writes:

> There is no issue with libffi from latest GIT compiled with
> llvm-gcc-4.2, and guile-2.0.5 compiled with SVN gcc-4.7.

But there is an issue with libffi from git compiled with llvm-gcc-4.2,
and guile-2.0.5 compiled with llvm-gcc-4.2?

Can you try compiling libffi from GIT with gcc-4.7, and guile-2.0.5 with
llvm-gcc-4.2?

Just to check :)

Andy
-- 
http://wingolog.org/

bug#10681: GNU Guile 2.0.5 released

[Hans Aberg]

On 1 Feb 2012, at 15:53, Andy Wingo wrote:

>> There is no issue with libffi from latest GIT compiled with
>> llvm-gcc-4.2, and guile-2.0.5 compiled with SVN gcc-4.7.
> 
> But there is an issue with libffi from git compiled with llvm-gcc-4.2,
> and guile-2.0.5 compiled with llvm-gcc-4.2?

Right, only that I think the issue is with guile-2.0.5 compiled with llvm-gcc-4.2.

> Can you try compiling libffi from GIT with gcc-4.7, and guile-2.0.5 with
> llvm-gcc-4.2?
> 
> Just to check :)

Unfortunately, llvm-gcc-4.2 is very slow; compiling guile-2.0.5 takes a very long time. So I think I will have to give up on this one.

Hans

bug#10681: GNU Guile 2.0.5 released

[Andy Wingo]

On Wed 01 Feb 2012 16:08, Hans Aberg <haberg-1@telia.com> writes:

> On 1 Feb 2012, at 15:53, Andy Wingo wrote:
>
>>> There is no issue with libffi from latest GIT compiled with
>>> llvm-gcc-4.2, and guile-2.0.5 compiled with SVN gcc-4.7.
>> 
>> But there is an issue with libffi from git compiled with llvm-gcc-4.2,
>> and guile-2.0.5 compiled with llvm-gcc-4.2?
>
> Right, only that I think the issue is with guile-2.0.5 compiled with llvm-gcc-4.2.
>
>> Can you try compiling libffi from GIT with gcc-4.7, and guile-2.0.5 with
>> llvm-gcc-4.2?
>> 
>> Just to check :)
>
> Unfortunately, llvm-gcc-4.2 is very slow; compiling guile-2.0.5 takes a very long time. So I think I will have to give up on this one.

Closing this one as done then.  Whenever you give a newer Guile a try
(like tomorrow's 2.0.6), we can look again.

Thanks!

Andy
-- 
http://wingolog.org/

bug#10681: GNU Guile 2.0.6 released

[Hans Aberg]

On 6 Jul 2012, at 20:23, Andy Wingo wrote:

> Closing this one as done then.  Whenever you give a newer Guile a try
> (like tomorrow's 2.0.6), we can look again.

Two tests failed. Drop me a note if you want further investigation.

Hans


----
$ make check
  GEN      public-submodule-commit
make  check-recursive
Making check in lib
make  check-recursive
make[4]: Nothing to be done for `check-am'.
Making check in meta
make[2]: Nothing to be done for `check'.
Making check in libguile
make  check-am
make[3]: Nothing to be done for `check-am'.
Making check in module
make[2]: Nothing to be done for `check'.
Making check in guile-readline
make  check-am
make[3]: Nothing to be done for `check-am'.
Making check in examples
make[2]: Nothing to be done for `check'.
Making check in emacs
make[2]: Nothing to be done for `check'.
Making check in test-suite
Making check in standalone
make  check-am
make  test-num2integral test-round test-list test-unwind test-conversion test-loose-ends test-scm-c-read test-scm-take-locale-symbol test-scm-take-u8vector test-scm-to-latin1-string test-scm-values test-with-guile-module test-scm-with-guile test-scm-spawn-thread test-pthread-create test-pthread-create-secondary test-system-cmds test-bad-identifiers test-require-extension test-guile-snarf test-import-order test-command-line-encoding test-command-line-encoding2 test-asmobs test-ffi test-fast-slot-ref test-mb-regexp test-use-srfi test-extensions
  CC       test_num2integral-test-num2integral.o
  CCLD     test-num2integral
  CC       test_round-test-round.o
  CCLD     test-round
  CC       test_list-test-list.o
  CCLD     test-list
  CC       test_unwind-test-unwind.o
  CCLD     test-unwind
  CC       test_conversion-test-conversion.o
  CCLD     test-conversion
  CC       test_loose_ends-test-loose-ends.o
  CCLD     test-loose-ends
  CC       test_scm_c_read-test-scm-c-read.o
  CCLD     test-scm-c-read
  CC       test_scm_take_locale_symbol-test-scm-take-locale-symbol.o
  CCLD     test-scm-take-locale-symbol
  CC       test_scm_take_u8vector-test-scm-take-u8vector.o
  CCLD     test-scm-take-u8vector
  CC       test_scm_to_latin1_string-test-scm-to-latin1-string.o
  CCLD     test-scm-to-latin1-string
  CC       test_scm_values-test-scm-values.o
  CCLD     test-scm-values
  CC       test_with_guile_module-test-with-guile-module.o
  CCLD     test-with-guile-module
  CC       test_scm_with_guile-test-scm-with-guile.o
  CCLD     test-scm-with-guile
  CC       test_scm_spawn_thread-test-scm-spawn-thread.o
  CCLD     test-scm-spawn-thread
  CC       test_pthread_create-test-pthread-create.o
  CCLD     test-pthread-create
  CC       test_pthread_create_secondary-test-pthread-create-secondary.o
  CCLD     test-pthread-create-secondary
make[5]: Nothing to be done for `../../../guile-2.0.6/test-suite/standalone/test-system-cmds'.
make[5]: Nothing to be done for `../../../guile-2.0.6/test-suite/standalone/test-bad-identifiers'.
make[5]: Nothing to be done for `../../../guile-2.0.6/test-suite/standalone/test-require-extension'.
make[5]: Nothing to be done for `../../../guile-2.0.6/test-suite/standalone/test-guile-snarf'.
make[5]: Nothing to be done for `../../../guile-2.0.6/test-suite/standalone/test-import-order'.
make[5]: Nothing to be done for `../../../guile-2.0.6/test-suite/standalone/test-command-line-encoding'.
make[5]: Nothing to be done for `../../../guile-2.0.6/test-suite/standalone/test-command-line-encoding2'.
make[5]: Nothing to be done for `../../../guile-2.0.6/test-suite/standalone/test-asmobs'.
make[5]: Nothing to be done for `../../../guile-2.0.6/test-suite/standalone/test-ffi'.
make[5]: `test-fast-slot-ref' is up to date.
make[5]: Nothing to be done for `../../../guile-2.0.6/test-suite/standalone/test-mb-regexp'.
make[5]: `test-use-srfi' is up to date.
make[5]: Nothing to be done for `../../../guile-2.0.6/test-suite/standalone/test-extensions'.
make  check-TESTS
PASS: test-system-cmds
PASS: test-bad-identifiers
PASS: test-require-extension
PASS: test-guile-snarf
PASS: test-import-order
PASS: test-command-line-encoding
PASS: test-command-line-encoding2
PASS: test-num2integral
PASS: test-round
PASS: test-asmobs
bad return from expression `(f-sum -1 2000 -30000 40000000000)': expected 39999971999; got 39999972255
FAIL: test-ffi
PASS: test-list
PASS: test-unwind
fail: scm_is_unsigned_integer ((- (expt 2 64) 1), 0, 18446744073709551615) == 1
FAIL: test-conversion
PASS: test-loose-ends
PASS: test-fast-slot-ref
PASS: test-mb-regexp
PASS: test-use-srfi
PASS: test-scm-c-read
PASS: test-scm-take-locale-symbol
PASS: test-scm-take-u8vector
PASS: test-scm-to-latin1-string
PASS: test-scm-values
PASS: test-extensions
PASS: test-with-guile-module
PASS: test-scm-with-guile
PASS: test-scm-spawn-thread
PASS: test-pthread-create
PASS: test-pthread-create-secondary
==================================
2 of 29 tests failed
Please report to bug-guile@gnu.org
==================================
make[5]: *** [check-TESTS] Error 1
make[4]: *** [check-am] Error 2
make[3]: *** [check] Error 2
make[2]: *** [check-recursive] Error 1
make[1]: *** [check-recursive] Error 1
make: *** [check] Error 2
----