Bug #100467 for App-cpanminus-reporter: App-cpanminus-reporter does not send UNKNOWN results for failed compilations

Thu Nov 20 01:51:27 2014 SREZIC [...] cpan.org - Ticket created

Subject:

App-cpanminus-reporter does not send UNKNOWN results for failed compilations

In the case of failed compilations CPAN::Reporter sends UNKNOWN reports, where App::cpanminus::reporter makes this into a FAIL. Here's an example: * http://www.cpantesters.org/cpan/report/d937eebc-5616-11e4-acef-b75a82024ea0 * http://www.cpantesters.org/cpan/report/f242119a-703c-11e4-83cf-8272e0bfc7aa Until this is fixed I urge every tester to NOT use App::cpanminus::reporter anymore. I marked this as critical because sending inconsistent results makes analysis (e.g. using http://analysis.cpantesters.org) difficult.

Thu Nov 20 13:07:02 2014 ether [...] cpan.org - Cc ETHER added

Sat Jan 31 19:26:04 2015 Marcin.Kasperski [...] mekk.waw.pl - Correspondence added

To add some research to this: cpanm-reporter at the moment uses cpanm installation status. CPAN::Reporter seems to have various heuristics inside. In this very case, CPAN/Reporter.pm has its _compute_make_grade routine, which makes "unknown" out of anything with non-zero exit value. For cpanm-reporter it is more difficult to know that make was used and that it failed with some status, as information must be parsed out, and actual text is volatile. Here is how the relevant part looks on my machine (Polish locale): ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~` cc -c -I/usr/lib/perl5/EV -I/usr/lib/perl5 -D_REENTRANT -D_GNU_SOURCE -DDEBIAN -fstack-protector -fno-strict-aliasing -pipe -I/usr/local/include -D_LARGEFILE_SOURCE -D_FILE_OFFSET_BITS=64 -O2 -g -DVERSION=\"2.2\" -DXS_VERSION=\"2.2\" -fPIC "-I/usr/lib/perl/5.18/CORE" ADNS.c ADNS.xs:6:18: fatal error: adns.h: Nie ma takiego pliku ani katalogu #include <adns.h> ^ compilation terminated. make: *** [ADNS.o] Błąd 1 -> FAIL Installing EV::ADNS failed. See /home/marcink/.cpanm/work/1422749568.24337/build.log for details. Retry with --force to force install it. ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Still, I suppose make: *** [...] [...] 1 present in the very last line before test FAIL is clear enough...

Sat Jan 31 19:26:05 2015 The RT System itself - Status changed from 'new' to 'open'

Sun Feb 01 02:30:26 2015 SREZIC [...] cpan.org - Correspondence added

On 2015-01-31 19:26:04, Mekk wrote: Show quoted text

> To add some research to this: cpanm-reporter at the moment uses cpanm > installation status. CPAN::Reporter seems to have various heuristics > inside. In this very case, CPAN/Reporter.pm has its > _compute_make_grade routine, which makes "unknown" out of anything > with non-zero exit value. > > For cpanm-reporter it is more difficult to know that make was used and > that it failed with some status, as information must be parsed out, > and actual text is volatile. Here is how the relevant part looks on my > machine (Polish locale): > > ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~` > cc -c -I/usr/lib/perl5/EV -I/usr/lib/perl5 -D_REENTRANT -D_GNU_SOURCE > -DDEBIAN -fstack-protector -fno-strict-aliasing -pipe > -I/usr/local/include -D_LARGEFILE_SOURCE -D_FILE_OFFSET_BITS=64 -O2 -g > -DVERSION=\"2.2\" -DXS_VERSION=\"2.2\" -fPIC "- > I/usr/lib/perl/5.18/CORE" ADNS.c > ADNS.xs:6:18: fatal error: adns.h: Nie ma takiego pliku ani katalogu > #include <adns.h> > ^ > compilation terminated. > make: *** [ADNS.o] Błąd 1 > -> FAIL Installing EV::ADNS failed. See > /home/marcink/.cpanm/work/1422749568.24337/build.log for details. > Retry with --force to force install it. > ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ > > > Still, I suppose > make: *** [...] [...] 1 > present in the very last line before test FAIL is clear enough...

Well, for GNU make. Different makes have different output here: - FreeBSD make: *** [fail] Error code 1 Stop in /tmp. - GNU make, English version: Makefile:2: recipe for target 'fail' failed gmake: *** [fail] Error 1 - GNU make, German version, latin1 locale: Makefile:2: die Regel für Ziel "fail" scheiterte gmake: *** [fail] Fehler 1 - GNU make, German version, utf8 locale (note the different quotes): Makefile:2: die Regel für Ziel „fail“ scheiterte gmake: *** [fail] Fehler 1 - dmake on FreeBSD: dmake: Error code 1, while making 'fail' At least nmake output should be also checked for, and are there any other popular makes?

Sun Feb 01 02:55:14 2015 SREZIC [...] cpan.org - Correspondence added

On 2015-02-01 02:30:26, SREZIC wrote: Show quoted text

> On 2015-01-31 19:26:04, Mekk wrote:

> > To add some research to this: cpanm-reporter at the moment uses cpanm > > installation status. CPAN::Reporter seems to have various heuristics > > inside. In this very case, CPAN/Reporter.pm has its > > _compute_make_grade routine, which makes "unknown" out of anything > > with non-zero exit value. > > > > For cpanm-reporter it is more difficult to know that make was used > > and > > that it failed with some status, as information must be parsed out, > > and actual text is volatile. Here is how the relevant part looks on > > my > > machine (Polish locale): > > > > ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~` > > cc -c -I/usr/lib/perl5/EV -I/usr/lib/perl5 -D_REENTRANT > > -D_GNU_SOURCE > > -DDEBIAN -fstack-protector -fno-strict-aliasing -pipe > > -I/usr/local/include -D_LARGEFILE_SOURCE -D_FILE_OFFSET_BITS=64 -O2 > > -g > > -DVERSION=\"2.2\" -DXS_VERSION=\"2.2\" -fPIC "- > > I/usr/lib/perl/5.18/CORE" ADNS.c > > ADNS.xs:6:18: fatal error: adns.h: Nie ma takiego pliku ani katalogu > > #include <adns.h> > > ^ > > compilation terminated. > > make: *** [ADNS.o] Błąd 1 > > -> FAIL Installing EV::ADNS failed. See > > /home/marcink/.cpanm/work/1422749568.24337/build.log for details. > > Retry with --force to force install it. > > ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ > > > > > > Still, I suppose > > make: *** [...] [...] 1 > > present in the very last line before test FAIL is clear enough...

> > Well, for GNU make. Different makes have different output here: > > - FreeBSD make: > *** [fail] Error code 1 > > Stop in /tmp. > > - GNU make, English version: > Makefile:2: recipe for target 'fail' failed > gmake: *** [fail] Error 1 > > - GNU make, German version, latin1 locale: > Makefile:2: die Regel für Ziel "fail" scheiterte > gmake: *** [fail] Fehler 1 > > - GNU make, German version, utf8 locale (note the different quotes): > Makefile:2: die Regel für Ziel „fail“ scheiterte > gmake: *** [fail] Fehler 1 > > - dmake on FreeBSD: > dmake: Error code 1, while making 'fail' > > At least nmake output should be also checked for, and are there any > other popular makes?

Another one, a make on Solaris: *** Error code 255 make: Fatal error: Command failed for target `fail'

Sat Apr 23 11:28:06 2016 garu [...] cpan.org - Correspondence added

Hi everyone, I'm sorry I didn't reply to this sooner, and I really appreciate your patience. cpanm-reporter is indeed unable to see the return value of make. However, we do rely on the output of cpanm and in this case, it looks like it's returning FAIL instead of UNKNOWN. Since this is mostly likely an error in the build.log parser, could you please send me the relevant "~/.cpanm/build.log" file so I can create test cases and fix the issue? Thanks!

Sat Dec 17 07:41:55 2016 twata_1 [...] yahoo.co.jp - Correspondence added

From:

twata_1 [...] yahoo.co.jp

Hi garu, Here is ~/.cpanm/build.log. ( Windows 7, Strawbery Perl v5.10.1 cpanm-reporter v0.17 ) Thnan you, -- twata On 2016-4月-23 土 11:28:06, GARU wrote: Show quoted text

> Hi everyone, > > I'm sorry I didn't reply to this sooner, and I really appreciate your > patience. > > cpanm-reporter is indeed unable to see the return value of make. > However, we do rely on the output of cpanm and in this case, it looks > like it's returning FAIL instead of UNKNOWN. Since this is mostly > likely an error in the build.log parser, could you please send me the > relevant "~/.cpanm/build.log" file so I can create test cases and fix > the issue? > > Thanks!

Subject:

build.log

Download build.log
application/octet-stream 2.8k

Message body not shown because it is not plain text.

Sat Apr 15 03:05:17 2017 SREZIC [...] cpan.org - Broken in 0.17 added

Wed May 31 10:40:14 2017 twata_1 [...] yahoo.co.jp - Correspondence added

From:

twata_1 [...] yahoo.co.jp

Hi garu, Here is other build logs. ( Windows 7, Strawbery Perl v5.10.1 cpanm-reporter v0.17 ) a dmake above: dmake.exe: Error code 129, while making (The rest is omitted.) Thank you, -- twata On 2016-12月-17 土 07:41:55, twata_1@yahoo.co.jp wrote: Show quoted text

> Hi garu, > > Here is ~/.cpanm/build.log. > ( > Windows 7, Strawbery Perl v5.10.1 > cpanm-reporter v0.17 > ) > > Thnan you, > > -- > twata > > > On 2016-4月-23 土 11:28:06, GARU wrote:

> > Hi everyone, > > > > I'm sorry I didn't reply to this sooner, and I really appreciate your > > patience. > > > > cpanm-reporter is indeed unable to see the return value of make. > > However, we do rely on the output of cpanm and in this case, it looks > > like it's returning FAIL instead of UNKNOWN. Since this is mostly > > likely an error in the build.log parser, could you please send me the > > relevant "~/.cpanm/build.log" file so I can create test cases and fix > > the issue? > > > > Thanks!

> >

Subject:

002-build.log

Download 002-build.log
application/octet-stream 11.5k

Message body not shown because it is not plain text.

Subject:

003-build.log

Download 003-build.log
application/octet-stream 35.8k

Message body not shown because it is not plain text.

Subject:

001-build.log

Download 001-build.log
application/octet-stream 30k

Message body not shown because it is not plain text.