[ Recipients of this message are either subscribers to the modperl or perlporters lists, or listed as authors of theinvolved modules]
> -----Original Message----- > From: [mailto:[EMAIL PROTECTED]] > Sent: Monday, June 03, 2002 9:38 PM > To: Alessandro Forghieri > Cc: [EMAIL PROTECTED] > Subject: Re: [2.x] test fails on win32 version 1.99_02 > > > On Monday, 2002-06-03 at 18:32:54 +0200, Alessandro Forghieri wrote: > > > reetings. > > :-) > > > I have been trying 5.8RC1 and I tried to compile/install > several modules > > that I have > > on 5.6.1 - CPAN, not PPM. > > At least the following do not build and/or test right (not > exhaustive, as I > > got tired after a while): > > > > Posted in the hope it may help somebody... > [EMAIL PROTECTED] Wrote: > Can you please send a detailed report to the perl5porters list? > [EMAIL PROTECTED] > and the module authors? Please include the version numbers > and all error messages. > And so it goes... I am changing the subject, to make it more descriptive, as this does not really relate to modperl. Machine is wintel NTSP6, VC++6 SP3. 1) Besides failures on various modules, I also witnessed (today) a SEGFAULT on the tests while building a debug version (curiously, the stack trace ends up in msvcrt - I would've expected msvcrtd). The crash happens during: ../lib/Tie/File/t/12_longfetch_rs...... It appears to wedge in Perl_NEWFOROP: #ifdef PL_OP_SLAB_ALLOC { LOOP *tmp; NewOp(1234,tmp,1,LOOP); Copy(loop,tmp,1,LOOP); //<====b0000m! FreeOp(loop); loop = tmp; } #else the "loop" structure has a few fishy fields: loop->op_nextop and loop->op_lastop both show up as expression that "Cannot be evaluated" by the debugger (but maybe that's normal). As far as I remember, this did not happen on a previous release build. All other tests complete sucessfully. 2) and now on to the modules failures. I trid to reapply my 5.6.1 AS630 bundle and I had several failures. I tracked three of them more closely. Two (Tk and Archive::Tar) are described below in a message I posted to perlmonks yesterday. The third - MLDBM is described in what follows. MLDBM-2.000 fails the first test of each suite. I debugged the first failure (dumper.t) and it fails a comparison on two would-be equal structures: DB<3> p $first $a = [ 1, { c => [ \'c' ], a => $a, b => $a->[1] }, $a->[1]{c} ]; $b = { c => [ \'c' ], a => [ 1, $b, $b->{c} ], b => $b }; $c = [ \'c' ]; DB<4> p $second $a = [ 1, { a => $a, b => $a->[1], c => [ \'c' ] }, $a->[1]{c} ]; $b = { a => [ 1, $b, [ \'c' ] ], b => $b, c => $b->{a}[2] }; $c = [ \'c' ]; So this is pretty serious - given the above interpreter error in a 'tie' test, some relation among the two can be inferred. I also had uninvestigated failures in the following modules - all at their latest CPAN versions: o) Devel::TraceLoad 0.08 # Test Failure t\all....FAILED test 1 Failed 1/5 tests, 80.00% okay Failed Test Stat Wstat Total Fail Failed List of Failed ---------------------------------------------------------------------------- --- t\all.t 5 1 20.00% 1 Failed 1/1 test scripts, 0.00% okay. 1/5 subtests failed, 80.00% okay. o) Module::Info 0.12 # Several test errors Test returned status 16 (wstat 4096, 0x1000) DIED. FAILED tests 10-15, 40-47, 50-51 Failed 16/51 tests, 68.63% okay Failed Test Stat Wstat Total Fail Failed List of Failed ---------------------------------------------------------------------------- --- t\Module-Info.t 16 4096 51 16 31.37% 10-15 40-47 50-51 Failed 1/1 test scripts, 0.00% okay. 16/51 subtests failed, 68.63% okay. o) Win32::API 0.20 #C diagnostic API.c API.xs(155) : error C2065: 'na' : undeclared identifier API.xs(237) : warning C4244: '=' : conversion from 'double ' to 'float ', possib le loss of data o) libxml-perl 0.07 # Fails test D:\perl58\bin\perl.exe "-MExtUtils::Command::MM" "-e" "test_harness(0, ' blib\lib', 'blib\arch')" t\amsterdam.t t\canon_xml_writer.t t\schema.t t\stream. t t\subs.t t\xp_sax.t t\amsterdam...........ok t\canon_xml_writer....ok t\schema..............ok t\stream..............FAILED test 11 Failed 1/11 tests, 90.91% okay t\subs................ok t\xp_sax..............ok Failed Test Stat Wstat Total Fail Failed List of Failed ---------------------------------------------------------------------------- --- t\stream.t 11 1 9.09% 11 Failed 1/6 test scripts, 83.33% okay. 1/45 subtests failed, 97.78% okay. o)XML-RSS 0.97 is uncorrectly packaged (it contains XML::RSS.3 - an illegal file name under windoze) and so unpack fails - but it builds/tests otherwise. o) Compress-zlib 1.1.16 The build process is broken with the latest zlib (1.1.4) and it requires extensive manual intervention (the build of zlib loops). Also the fact that building aganst a dll cannot be done (the XS extension and zlib proper are both called zlib.dll and the loader gets confused - something similar used to happen with expat) should be documented. o) libwin32-0.18 The test/install process is broken. The files: D:/builds/libwin32-0.18/APIFile/cFile.pc D:/builds/libwin32-0.18/APIRegistry/cRegistry.pc Are not copied to the final destination and tests fail (this also happened in 5.6.1) tests/installs OK otherwise. TK and Archive::Tar failures are documented below. Also attached is the autobundle documenting (so to speak) a number of modules that did build/test/install. Cheers, alf ----Posted to the perlmonks ---- Greetings, I have built 5.8.RC1 on my wintel NTSP6 box (VC++ 6 SP3 - smooth) I am now trying to reapply the bundle of what is already installed on my good ole 5.6.1 - and this is not going very smoothly. Until now I have been unable to build Archive::Tar and Tk (surely others will pop up). Archive::Tar errors like this: D:\home\alf\.cpan58\build\Archive-Tar-0.22>nmake test ... cp Tar.pm blib\lib\Archive\Tar.pm D:\perl58\bin\perl.exe -MExtUtils::Command -e cp ptar blib\script\ptar pl2bat.bat blib\script\ptar D:\perl58\bin\perl.exe "-Iblib\lib" "-Iblib\arch" test.pl 1..10 ok 1 ok 2 ok 3 Can't call method "gzwrite" on an undefined value at blib\lib/Archive/Tar.pm line 493. NMAKE : fatal error U1077: 'D:\perl58\bin\perl.exe' : return code '0xff' Stop. The error is related to this line: $fh = Compress::Zlib::gzdopen_ ($fh, $mode, 0) or &_drat; I tried to change it thusly: $fh = Compress::Zlib::gzdopen_ (fileno($fh),$mode,0) or &_drat; Which solves the problem in test 4 and 5, only to fail on test 6: ok 5 Can't call method "gzseek" on an undefined value at blib\lib/Archive/Tar.pm line 846. The reason for this failure is deeper however and I cannot comprehend it, also because it appears to be time-dependent: if I step through the code with the debugger quickly the failure appears rooted at the line: sysseek ($_[0], 0, 0) or goto &_drat; which 'drats', while if I debug more slowly i can get to the end of the test (which prints not ok 6) and the situation unravels later on; Tk gets a straight compile failure: cl -c -I.. -I../pTk/mTk/xlib -nologo -Gf -W3 -MD -DNDEBUG -O1 -DWIN32 -D_CONSOLE -DNO_STRICT -DHAVE_DES_FCRYPT -DPERL_IMPLICIT_CONTEXT -DPERL_IMPLICIT _SYS -DUSE_PERLIO -DPERL_MSVCRT_READFIX -MD -DNDEBUG -O1 -DVERSION=\"800.024\ " -DXS_VERSION=\"800.024\" "-ID:\perl58\lib\CORE" Xlib.c Xlib.c ..\pTk/tkWinPort.h(135) : error C2143: syntax error : missing ')' before '(' ..\pTk/tkWinPort.h(135) : error C2143: syntax error : missing ')' before '(' ..\pTk/tkWinPort.h(135) : error C2091: function returns function ..\pTk/tkWinPort.h(135) : error C2091: function returns function ..\pTk/tkWinPort.h(135) : error C2146: syntax error :missing ')' befor +e identifier 'Perl_get_context' NMAKE : fatal error U1077: 'cl' : return code '0x2' Google does not really help here so - opinions? Ideas? TIA & Cheers, alf
Snap.zip
Description: Binary data