GNU bug report logs - #45205
Failed tests for 1.16.3 on Solaris

Previous Next

To reply to this bug, email your comments to 45205 AT debbugs.gnu.org.

Toggle the display of automated, internal messages from the tracker.

Message #5 received at submit <at> debbugs.gnu.org (full text, mbox):

From: Dagobert Michelsen <dam <at> opencsw.org>
To: Ken Moffat via Bug reports for Automake <bug-automake <at> gnu.org>
Subject: Failed tests for 1.16.3 on Solaris
Date: Sat, 12 Dec 2020 13:23:51 +0100

[Message part 1 (text/plain, inline)]

Hi.

I have 14 failed tests for automake 1.16.3 on Solaris 10 Sparc. Please see
the attached log for details.

[test-suite.log.gz (application/x-gzip, attachment)]

[Message part 3 (text/plain, inline)]

Best regards

  — Dago

-- 
"You don't become great by trying to be great, you become great by wanting to do something,
and then doing it so hard that you become great in the process." - xkcd #896

Message #8 received at 45205 <at> debbugs.gnu.org (full text, mbox):

From: Karl Berry <karl <at> freefriends.org>
To: dam <at> opencsw.org
Cc: 45205 <at> debbugs.gnu.org
Subject: Re: bug#45205: Failed tests for 1.16.3 on Solaris
Date: Sun, 13 Dec 2020 16:14:30 -0700

Hi Dagobert - thanks for the report.

Regarding Solaris lex, and maybe other compiler tools - yes, the failure
is known (another bug with the same issue:
https://debbugs.gnu.org/cgi/bugreport.cgi?bug=44795). Unfortunately I
cannot debug, not having any access to this system, or those tools, or
enough available time to want to spend it dealing with proprietary tools
even given a system. I'm happy to install patches if you, or anyone, can
figure out the problem(s).

Regarding the python failures: apparently the failure is because a use
of $(...) slipped into the py-compile script. As we all know, that
construct is not supported by Solaris /bin/sh. I'll fix py-compile. You
can do it locally if you want in the meantime, it's just one obvious
line:
python_major=$($PYTHON -V 2>&1 | sed -e 's/.* //;s/\..*$//;1q')

BTW, current config.{guess,sub} assume $(...) -- a recent change. I
reported this as a problem but no reversion has been forthcoming. You
might or might not want to report it also. My original message:
https://lists.gnu.org/archive/html/config-patches/2020-12/msg00000.html

Regarding vala-mix2: this one might have been solved in the meantime.
Timestamp problem.

Regarding the other failures: I admit I cannot discern the problem just
from inspecting the log, nor can I debug at a distance. Sorry. Patches
welcome, as above. 

For what it's worth, the changes for 1.16.3 (and 1.16.2) were
minor. Nothing that knowingly broke Solaris or induced dependencies on
newer tools. Most of the failing tests were not touched. So I'm sadly
unclear as to what's going on.

I'm going to mark this bug as "confirmed" and also "needs help",
since I can't solve it :(.

All the best,
Karl

Message #13 received at 45205 <at> debbugs.gnu.org (full text, mbox):

From: Bogdan <bogdro_rep <at> gmx.us>
To: automake-patches <at> gnu.org, 34151 <at> debbugs.gnu.org, 42393 <at> debbugs.gnu.org,
 44795 <at> debbugs.gnu.org, 49755 <at> debbugs.gnu.org, 45205 <at> debbugs.gnu.org,
 55073 <at> debbugs.gnu.org
Subject: [PATCH][bug 34151, 42393, 44795, 49755, 45205, 55073] Fix lex/yacc
 C++ tests on Solaris/SunOS
Date: Sun, 16 Apr 2023 18:37:28 +0200

[Message part 1 (text/plain, inline)]

Hi all.

 The Solaris/SunOS lex generates C++ code with functions properly 
wrapped into >extern "C"<, but only if:

- compiling as C++ (__cplusplus is #defined),

- the feature constant __EXTERN_C__ is #defined (probably meant just 
for the SunOS compiler, not for all).

Other compilers, like GCC/G++, obviously miss the SunOS-specific 
__EXTERN_C__ and thus miss >extern "C"<, declaring the yylex() 
function as C++ (the default). This doesn't need to be a problem, but:

- the function should be declared as the user code expects it (as C 
for C programs and as C++ for C++ programs),

- the Automake lex tests for C++ re-declare the function always as a C 
function.

 The re-declarations happens probably for good reason, like for lex 
implementations which don't declare the function in advance combined 
with C/C++ compilers which always require declarations. Unfortunately, 
having the same function declared as C++ (default) first and then 
later as C causes compiler errors.

 The attached patch corrects the declarations in the following way:

- if we're in C++ mode and __EXTERN_C__ is declared (so, we're on a 
SunOS with their compiler), declare the function as C in C++ mode 
(i.e., use >extern "C"<), just like the SunOS lex does,

- if we're in C++ mode and __EXTERN_C__ is NOT declared, but also 
"__sun" is not declared (so, we're NOT on a SunOS), still assume C++ 
mode, but a different system and declare the function as C in C++ mode 
(i.e., use >extern "C"<) to support other compilers,

- if we're in C++ mode and __EXTERN_C__ is NOT declared, but also 
"__sun" IS declared (so, we ARE on a SunOS, but not using their 
compiler), declare the function as C++ in C++ mode, just like the 
SunOS lex does,

- if we're NOT in C++ mode, declare the function as C. Not the case in 
these tests, though.

 This should match what the SunOS lex/compiler do and expect which 
not braking other systems (re-checked on my Linux - works).

 Verified that the defect exists on a Solaris 11.4 Build 15 (SunOS 
5.11, GCC 7.3.0, amd64) and that the patch fixes it.

 Furthermore, I checked how GCC manages the declarations of standard 
C functions in C++. Seems so simple as saying:

	using std::some_c_library_function_name;

 which the SunOS library (or maybe just the old GCC library there) is 
not doing.
 So, it just takes this to fix the missing prototypes in C++ mode.

 In short, the patch:
- fixes the remaining part of 
https://debbugs.gnu.org/cgi/bugreport.cgi?bug=34151,
- fixes the remaining part of 
https://debbugs.gnu.org/cgi/bugreport.cgi?bug=42393 
(t/silent-many-languages.sh works for me),
- fixes https://debbugs.gnu.org/cgi/bugreport.cgi?bug=44795,
- fixes https://debbugs.gnu.org/cgi/bugreport.cgi?bug=49755,
- partially fixes https://debbugs.gnu.org/cgi/bugreport.cgi?bug=45205,
- partially fixes https://debbugs.gnu.org/cgi/bugreport.cgi?bug=55073.

 Fortunately, these parts are just problems with the tests 
themselves, not with Automake in general.

-- 
Regards - Bogdan ('bogdro') D.                 (GNU/Linux & FreeDOS)
X86 assembly (DOS, GNU/Linux):    http://bogdro.evai.pl/index-en.php
Soft(EN): http://bogdro.evai.pl/soft  http://bogdro.evai.pl/soft4asm
www.Xiph.org  www.TorProject.org  www.LibreOffice.org  www.GnuPG.org

[automake-sunos2-mail.diff (text/x-patch, attachment)]

Message #16 received at 45205 <at> debbugs.gnu.org (full text, mbox):

From: Bogdan <bogdro_rep <at> gmx.us>
To: 45205 <at> debbugs.gnu.org, 55073 <at> debbugs.gnu.org
Subject: GNU Automake - testsuite failures on Solaris
Date: Sun, 16 Apr 2023 18:40:13 +0200

Hi.

 Failures in lex/yacc/C++ tests should be fixed now with the patch in 
my previous mail.
 I'm afraid the remaining test failures may indeed point to an actual 
bug in Automake, not just in tests.

 Regarding failures in tests:
- subpkg,
- subpkg2,
- subpkg-yacc,
as far as I've investigated, the problem is that $(distdir) = 
$(top_distdir) in a subpackage while this apparently should never be 
the case. This makes the file checks:

	test ! -f $(distdir)/LDADD.c
	test -f $(top_distdir)/LDADD.c

fail for the obvious reason - if the directory is the same, then 
either the file exists, or it doesn't, so one of the tests will always 
fail.
 No idea why the directories are the same, though.

 As for fn99subdir: make distcheck fails, because the 'cnfsubdir' dir 
doesn't exist in the distribution directory/package while it 
apparently should. No idea why the directory is not packaged, but at 
lease I know it's not a problem of e.g. the length of the path (what I 
suspected earlier).

 As for subpkg3: the subpkg directory exists, but doesn't contain 
'configure'. Also no idea why.

 There must be something system-specific, obviously, that makes 
Automake behave not the way it's supposed to. Sure, it can also be a 
problem in the tests themselves, but I don't see why they would 
succeed on a Linux and fail on a Solaris/SunOS.

-- 
Regards - Bogdan ('bogdro') D.                 (GNU/Linux & FreeDOS)
X86 assembly (DOS, GNU/Linux):    http://bogdro.evai.pl/index-en.php
Soft(EN): http://bogdro.evai.pl/soft  http://bogdro.evai.pl/soft4asm
www.Xiph.org  www.TorProject.org  www.LibreOffice.org  www.GnuPG.org

Message #19 received at 45205 <at> debbugs.gnu.org (full text, mbox):

From: Karl Berry <karl <at> freefriends.org>
To: dclarke <at> blastwave.org, dam <at> opencsw.org, yoi_no_myoujou <at> yahoo.co.jp
Cc: 42393 <at> debbugs.gnu.org, 45205 <at> debbugs.gnu.org, 44795 <at> debbugs.gnu.org,
 30612 <at> debbugs.gnu.org, 31126 <at> debbugs.gnu.org, 55073 <at> debbugs.gnu.org,
 49755 <at> debbugs.gnu.org, 15256 <at> debbugs.gnu.org, 34151 <at> debbugs.gnu.org,
 bogdro_rep <at> gmx.us
Subject: Re: bug#31126: GNU Automake 1.16.1 fails 21 tests on Solaris 10
 sparcv9
Date: Sun, 9 Jul 2023 10:57:42 -0600

Hi - you have each reported Automake test failures on Solaris.

I've just pushed a fix from Bogdan (thanks Bogdan!) for some of these,
which (sometimes with other fixes already made) closes some bugs.
Bogdan explains the details of the latest:
https://debbugs.gnu.org/cgi/bugreport.cgi?bug=45205#13

Other problems remain and still need debugging, as Bogdan explains in:
https://debbugs.gnu.org/cgi/bugreport.cgi?bug=45205#16

If anyone has time and interest to try bootstrapping current automake,
run the test suite, and find what still fails and why, that would be
great. --thanks, karl.

Bugs closed (with fingers crossed): 15256 30612 34151 42393 44795 49755.
Bugs still open: 31126 45205 55073.


2023-07-09  Bogdan  <bogdro_rep <at> gmx.us>

tests: avoid some SunOS declaration conflicts.

This change: finishes fixing https://bugs.gnu.org/34151
  and https://bugs.gnu.org/30612 and https://bugs.gnu.org/42393;
fixes https://bugs.gnu.org/44795 and https://bugs.gnu.org/49755;
partially fixes https://bugs.gnu.org/45205
  and https://bugs.gnu.org/55073.
Still open: https://bugs.gnu.org/15256;

* t/lex-clean-cxx.sh: test __cplusplus, __sun, and __EXTERN_C__
before using extern "C" for the yylex decl, as explained at
https://debbugs.gnu.org/cgi/bugreport.cgi?bug=45205#13.
* t/lex-depend-cxx.sh: likewise.

* t/yacc-cxx.sh: "using" std::exit, std::free, std::malloc.
* t/yacc-d-cxx.sh: likewise.
* t/yacc-mix-c-cxx.sh: likewise.

* NEWS: mention this.
diff --git a/NEWS b/NEWS
index 61d631ba0..23b2cc50a 100644
--- a/NEWS
+++ b/NEWS
@@ -46,6 +46,8 @@ New in 1.17:
   - distcleancheck ignores "silly rename" files (.nfs* .smb* .__afs*)
     that can show up on network file systems.

+  - tests: avoid some declaration conflicts for lex et al. on SunOS.
+
 * Obsolescent features:

   - py-compile no longer supports Python 0.x or 1.x versions.  Python 2.0,
diff --git a/t/lex-clean-cxx.sh b/t/lex-clean-cxx.sh
index 3632d5c87..8a086304f 100644
--- a/t/lex-clean-cxx.sh
+++ b/t/lex-clean-cxx.sh
@@ -51,10 +51,15 @@ CLEANFILES = parsebaz.l++ parsequx.lpp
 LDADD = $(LEXLIB)
 END

+# For the explanation of the conditionals on using extern "C",
+# see https://debbugs.gnu.org/cgi/bugreport.cgi?bug=45205#13.
 cat > parsefoo.lxx << 'END'
 %{
 #define YY_DECL int yylex (void)
-extern "C" YY_DECL;
+#if (defined __cplusplus) && ((!defined __sun) || (defined __EXTERN_C__))
+extern "C"
+#endif
+YY_DECL;
 #define YY_NO_UNISTD_H 1
 int isatty (int fd) { return 0; }
 %}
@@ -71,7 +76,10 @@ cp parsefoo.lxx parsebar.ll

 cat > mainfoo.cc << 'END'
 // This file should contain valid C++ but invalid C.
-extern "C" int yylex (void);
+#if (defined __cplusplus) && ((!defined __sun) || (defined __EXTERN_C__))
+extern "C"
+#endif
+int yylex (void);
 using namespace std;
 int main (int argc, char **argv)
 {
diff --git a/t/lex-depend-cxx.sh b/t/lex-depend-cxx.sh
index 60615a54e..682b2ccc5 100644
--- a/t/lex-depend-cxx.sh
+++ b/t/lex-depend-cxx.sh
@@ -44,10 +44,15 @@ test-obj-updated: joe.$(OBJEXT) moe.$(OBJEXT)
 	is_newest moe.$(OBJEXT) my-hdr.hxx
 END

+# For the explanation of the conditionals on using extern "C",
+# see https://debbugs.gnu.org/cgi/bugreport.cgi?bug=45205#13.
 cat > joe.ll << 'END'
 %{
 #define YY_DECL int yylex (void)
-extern "C" YY_DECL;
+#if (defined __cplusplus) && ((!defined __sun) || (defined __EXTERN_C__))
+extern "C"
+#endif
+YY_DECL;
 #define YY_NO_UNISTD_H 1
 int isatty (int fd) { return 0; }
 %}
diff --git a/t/yacc-cxx.sh b/t/yacc-cxx.sh
index 1bf75a975..5b385870c 100644
--- a/t/yacc-cxx.sh
+++ b/t/yacc-cxx.sh
@@ -49,6 +49,9 @@ cat > parse1.yy << 'END'
 // Valid C++, but deliberately invalid C.
 #include <cstdio>
 #include <cstdlib>
+using std::exit;
+using std::free;
+using std::malloc;
 // "std::" qualification required by Sun C++ 5.9.
 int yylex (void) { return std::getchar (); }
 void yyerror (const char *s) {}
diff --git a/t/yacc-d-cxx.sh b/t/yacc-d-cxx.sh
index 56f43c940..576e8e2aa 100644
--- a/t/yacc-d-cxx.sh
+++ b/t/yacc-d-cxx.sh
@@ -31,6 +31,9 @@ write_parse ()
     #include <stdlib.h>
     // Valid C++, but deliberately invalid C.
     #include <cstdlib>
+    using std::exit;
+    using std::free;
+    using std::malloc;
     #include "$header"
     int yylex (void) { return 0; }
     void yyerror (const char *s) {}
diff --git a/t/yacc-mix-c-cxx.sh b/t/yacc-mix-c-cxx.sh
index a73935c39..7c44e0926 100644
--- a/t/yacc-mix-c-cxx.sh
+++ b/t/yacc-mix-c-cxx.sh
@@ -86,6 +86,9 @@ cat > parse.yy <<'END'
 // https://bugs.gnu.org/20031
 #include <stdlib.h>
 // Valid C++, but deliberately invalid C.
+using std::exit;
+using std::free;
+using std::malloc;
 #include <cstdlib>
 #include "parse.hh"
 int yylex (void) { return 0; }

compile finished at Sun Jul  9 09:56:41 2023

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.