Bug #128950 for Test-Smoke: Branch selected was either not smoked or reported under older branch's name

Wed Mar 27 16:26:55 2019 jkeenan [...] pobox.com - Ticket created

Subject:	Branch selected was either not smoked or reported under older branch's name
Date:	Wed, 27 Mar 2019 16:26:40 -0400
To:	bug-Test-Smoke [...] rt.cpan.org
From:	James E Keenan <jkeenan [...] pobox.com>

Today on a FreeBSD-11.2 host from which I have been submitting smoke tests of the Perl 5 core distribution for at least eight months I had a situation where the branch I designated for smoking was not smoked; another branch was smoked instead. When I smoke a branch other than blead I simply type the name of the non-blead branch into the first record in 'smokecurrent.gitbranch', thereby pushing all other records in that file down one line. At 14:14 UTC I modified 'smokecurrent.gitbranch' to place 'smoke-me/khw-locale' in the first position. ##### [p5smoke] $ l smokecurrent.gitbranch -rw-r--r-- 1 jkeenan jkeenan 2140 Mar 27 14:14 smokecurrent.gitbranch [p5smoke] $ head -n 5 smokecurrent.gitbranch blead smoke-me/jkeenan/ntyni/133928-posix-2nd blead smoke-me/jkeenan/ntyni/133928-posix ##### I then resumed a screen session in which I called 'sh ./smokecurrent.sh'. This particular smoke configuration is quite thorough, so I did not check on its progress for the next 5 hours. When I checked just now, I discovered to my surprise that 'smoke-me/khw-locale' had not been smoked. Instead, 'smoke-me/jkeenan/ntyni/133928-posix-2nd' was smoked -- despite being the *third* record in smokecurrent.gitbranch. (I originally smoked this branch on March 16; see http://perl5.test-smoke.org/report/82582. All smoke runs between March 16 and March 26 were done on blead.) I'm attaching a file, 'febb9d45.freebsd-11.20190327.smokecurrent.reduced.log', which is smokecurrent.log with the data from the individual smoke passes stripped out. (They were all PASS, so the results are not at issue here.) In that file, note that, as expected, 'smoke-me/khw-locale' is the branch cited at lines 3, 6, 10 and 16, as expected. At line 19, 'febb9d45da2878c49f62aae8212fa4d4c13d883c' is cited as the 'patchlevel' in the 'perl-current/' directory. Commit febb9d45 was HEAD of the 'smoke-me/khw-locale' branch at that point in time. This commit cited again at line 29. However, at line 30 we suddenly see a different, older branch being cited: "Smoking branch smoke-me/jkeenan/ntyni/133928-posix-2nd". That branch is what gets cited in the web report (http://perl5.test-smoke.org/report/83471), even though febb9d45 is the SHA cited. Checking on the two repositories which are part of this Test-Smoke rig, I see that 'git-perl' is in the intended branch but 'perl-current' is in the non-intended, older branch: ##### [git-perl] $ gitcurr smoke-me/khw-locale [git-perl] $ cd ../perl-current/ [perl-current] $ gitcurr 'smoke-me/jkeenan/ntyni/133928-posix-2nd ##### ('gitcurr' is an alias for 'git branch' grepped for the current branch -- that with the asterisk.) However, when I call 'git show --format=fuller' in each of these two repositories, I get exactly the same output! (See attachments.) Additional data: The name of the requested branch shows up in the *.log file for this run. The name of the non-requested branch shows up in the *.jsn, *.log, *.out and *.rpt files for this run. ##### [smokecurrent] $ grep -l -E 'khw' *febb9d45* logfebb9d45da2878c49f62aae8212fa4d4c13d883c.log [smokecurrent] $ grep -l -E 'ntyni' *febb9d45* jsnfebb9d45da2878c49f62aae8212fa4d4c13d883c.jsn logfebb9d45da2878c49f62aae8212fa4d4c13d883c.log outfebb9d45da2878c49f62aae8212fa4d4c13d883c.out rptfebb9d45da2878c49f62aae8212fa4d4c13d883c.rpt ##### It is possible that the branch I requested to be smoked, 'smoke-me/khw-locale', was indeed smoked -- but if so it was ultimately recorded as being the result of a smoke of the 'smoke-me/jkeenan/ntyni/133928-posix-2nd' branch. And it is under the latter, non-intended name by which this smoke-test result is catalogued at http://perl5.test-smoke.org/search. So, either I've wasted 4+ hours of server time smoking the code of an older branch, or I've spent 4+ hours testing the correct branch but it's not being reported under that branch's name. Which is it? Can you explain why, starting at line 30 of the log file, the wrong branch name is being used? Thank you very much. Jim Keenan

Message body is not shown because sender requested not to inline it.

Thu May 16 21:38:53 2019 jkeenan [...] cpan.org - Correspondence added

On Wed Mar 27 16:26:55 2019, jkeenan@pobox.com wrote: Show quoted text

> Today on a FreeBSD-11.2 host from which I have been submitting smoke > tests of the Perl 5 core distribution for at least eight months I had a > situation where the branch I designated for smoking was not smoked; > another branch was smoked instead. > > When I smoke a branch other than blead I simply type the name of the > non-blead branch into the first record in 'smokecurrent.gitbranch', > thereby pushing all other records in that file down one line. > > At 14:14 UTC I modified 'smokecurrent.gitbranch' to place > 'smoke-me/khw-locale' in the first position. > > ##### > [p5smoke] $ l smokecurrent.gitbranch > -rw-r--r-- 1 jkeenan jkeenan 2140 Mar 27 14:14 smokecurrent.gitbranch > > [p5smoke] $ head -n 5 smokecurrent.gitbranch > blead > smoke-me/jkeenan/ntyni/133928-posix-2nd > blead > smoke-me/jkeenan/ntyni/133928-posix > ##### > > I then resumed a screen session in which I called 'sh > ./smokecurrent.sh'. This particular smoke configuration is quite > thorough, so I did not check on its progress for the next 5 hours. When > I checked just now, I discovered to my surprise that > 'smoke-me/khw-locale' had not been smoked. Instead, > 'smoke-me/jkeenan/ntyni/133928-posix-2nd' was smoked -- despite being > the *third* record in smokecurrent.gitbranch. (I originally smoked this > branch on March 16; see http://perl5.test-smoke.org/report/82582. All > smoke runs between March 16 and March 26 were done on blead.) > > I'm attaching a file, > 'febb9d45.freebsd-11.20190327.smokecurrent.reduced.log', which is > smokecurrent.log with the data from the individual smoke passes stripped > out. (They were all PASS, so the results are not at issue here.) > > In that file, note that, as expected, 'smoke-me/khw-locale' is the > branch cited at lines 3, 6, 10 and 16, as expected. At line 19, > 'febb9d45da2878c49f62aae8212fa4d4c13d883c' is cited as the 'patchlevel' > in the 'perl-current/' directory. Commit febb9d45 was HEAD of the > 'smoke-me/khw-locale' branch at that point in time. This commit cited > again at line 29. > > However, at line 30 we suddenly see a different, older branch being > cited: "Smoking branch smoke-me/jkeenan/ntyni/133928-posix-2nd". That > branch is what gets cited in the web report > (http://perl5.test-smoke.org/report/83471), even though febb9d45 is the > SHA cited. > > Checking on the two repositories which are part of this Test-Smoke rig, > I see that 'git-perl' is in the intended branch but 'perl-current' is in > the non-intended, older branch: > > ##### > [git-perl] $ gitcurr > smoke-me/khw-locale > [git-perl] $ cd ../perl-current/ > [perl-current] $ gitcurr > 'smoke-me/jkeenan/ntyni/133928-posix-2nd > ##### > > ('gitcurr' is an alias for 'git branch' grepped for the current branch > -- that with the asterisk.) > > However, when I call 'git show --format=fuller' in each of these two > repositories, I get exactly the same output! (See attachments.) > > Additional data: The name of the requested branch shows up in the *.log > file for this run. The name of the non-requested branch shows up in the > *.jsn, *.log, *.out and *.rpt files for this run. > > ##### > [smokecurrent] $ grep -l -E 'khw' *febb9d45* > logfebb9d45da2878c49f62aae8212fa4d4c13d883c.log > [smokecurrent] $ grep -l -E 'ntyni' *febb9d45* > jsnfebb9d45da2878c49f62aae8212fa4d4c13d883c.jsn > logfebb9d45da2878c49f62aae8212fa4d4c13d883c.log > outfebb9d45da2878c49f62aae8212fa4d4c13d883c.out > rptfebb9d45da2878c49f62aae8212fa4d4c13d883c.rpt > ##### > > It is possible that the branch I requested to be smoked, > 'smoke-me/khw-locale', was indeed smoked -- but if so it was ultimately > recorded as being the result of a smoke of the > 'smoke-me/jkeenan/ntyni/133928-posix-2nd' branch. And it is under the > latter, non-intended name by which this smoke-test result is catalogued > at http://perl5.test-smoke.org/search. >

I had another instance of this problem tonight. I believe that the branch I intended to smoke (as set in smokecurrent.gitbranch) was actually smoked, but it was reported as a different branch. See below. Show quoted text

> So, either I've wasted 4+ hours of server time smoking the code of an > older branch, or I've spent 4+ hours testing the correct branch but it's > not being reported under that branch's name. Which is it? > > Can you explain why, starting at line 30 of the log file, the wrong > branch name is being used? > > Thank you very much. > Jim Keenan

Tonight I intended to smoke this branch: smoke-me/khw-petr Here is the head of smokecurrent.log. ##### [2019-05-16 22:10:32+0000] ==> Starting synctree [2019-05-16 22:10:32+0000] qx[/usr/local/bin/git remote update --prune 2>&1] [2019-05-16 22:10:36+0000] qx[/usr/local/bin/git checkout smoke-me/khw-petr [2019-05-16 22:10:36+0000] 2>&1] Switched to a new branch 'smoke-me/khw-petr' [2019-05-16 22:10:36+0000] qx[/usr/local/bin/git reset --hard origin/smoke-me/khw-petr [2019-05-16 22:10:36+0000] 2>&1] [2019-05-16 22:10:36+0000] qx[/usr/local/bin/git clean -dfx 2>&1] [2019-05-16 22:10:37+0000] qx[/usr/local/bin/git fetch origin 2>&1] [2019-05-16 22:10:39+0000] qx[/usr/local/bin/git checkout smoke-me/khw-petr [2019-05-16 22:10:39+0000] 2>&1] error: Your local changes to the following files would be overwritten by checkout: patchlevel.h Please commit your changes or stash them before you switch branches. Aborting [2019-05-16 22:10:39+0000] qx[/usr/local/bin/git reset --hard origin/smoke-me/khw-petr [2019-05-16 22:10:39+0000] 2>&1] [2019-05-16 22:10:39+0000] qx[/usr/local/bin/perl5 Porting/make_dot_patch.pl > .patch] [2019-05-16 22:10:39+0000] /home/jkeenan/p5smoke/perl-current is now up to patchlevel 130031a372ca43b9e664276be0cb4986f39bc432 [2019-05-16 22:10:39+0000] ==> Starting runsmoke [2019-05-16 22:10:39+0000] [./tssmokeperl.pl] chdir(/home/jkeenan/p5smoke/perl-current) [2019-05-16 22:10:39+0000] Setup alarm: Fri May 17 07:10:39 2019 "x.log" 209L, 12600C 1,1 Top [2019-05-16 22:10:36+0000] qx[/usr/local/bin/git checkout smoke-me/khw-petr [2019-05-16 22:10:36+0000] qx[/usr/local/bin/git reset --hard origin/smoke-me/khw-petr [2019-05-16 22:10:36+0000] 2>&1] [2019-05-16 22:10:36+0000] qx[/usr/local/bin/git clean -dfx 2>&1] [2019-05-16 22:10:39+0000] 2>&1] Please commit your changes or stash them before you switch branches. [2019-05-16 22:10:39+0000] 2>&1] [2019-05-16 22:10:39+0000] ==> Starting runsmoke [2019-05-16 22:10:39+0000] [./tssmokeperl.pl] chdir(/home/jkeenan/p5smoke/perl-current) [2019-05-16 22:10:39+0000] Setup alarm: Fri May 17 07:10:39 2019 [2019-05-16 22:10:39+0000] qx[/usr/local/bin/perl5 -e "require q[/home/jkeenan/p5smoke/perl-current/cpan/Test-Harness/lib/Test/Harness.pm];print Test::Harness->VERSION" 2>&1] [2019-05-16 22:10:39+0000] Found: Test::Harness version 3.42. [2019-05-16 22:10:39+0000] Reading build configurations from /home/jkeenan/p5smoke/perlcurrent.cfg [2019-05-16 22:10:39+0000] Reading 'Policy.sh' from default content (v=1) [2019-05-16 22:10:39+0000] Running test suite only with 'harness' [2019-05-16 22:10:39+0000] make distclean ... [2019-05-16 22:10:39+0000] Smoking patch 130031a372ca43b9e664276be0cb4986f39bc432 v5.30.0-RC1-16-g130031a372 [2019-05-16 22:10:39+0000] Smoking branch blead ##### Note the last two lines above. 30031a372ca43b9e664276be0cb4986f39bc432 is indeed the SHA for HEAD in the smoke-me/khw-petr branch. But then in the last line it states that we're smoking 'blead'. And 'blead' is what got reported in http://perl5.test-smoke.org/report/87533! If I descend into the perl-current directory, I'm told that the current branch is 'blead'. ##### $ gitcurr blead ##### But 'git log' starts like this: ##### commit 130031a372ca43b9e664276be0cb4986f39bc432 (HEAD -> blead, origin/smoke-me/khw-petr) Author: Karl Williamson <khw@cpan.org> Date: Thu May 16 15:47:20 2019 -0600 [perl #134112] 5.30 regression in Unicode properties defined with splice ##### So the right SHA, the wrong branch, but the right "branch based on". Can you explain what is happening? Thank you very much. Jim Keenan

Thu May 16 21:38:54 2019 The RT System itself - Status changed from 'new' to 'open'

Thu May 16 21:53:15 2019 jkeenan [...] cpan.org - Correspondence added

On Thu May 16 21:38:53 2019, JKEENAN wrote: Show quoted text

> On Wed Mar 27 16:26:55 2019, jkeenan@pobox.com wrote:

> > Today on a FreeBSD-11.2 host from which I have been submitting smoke > > tests of the Perl 5 core distribution for at least eight months I had > > a > > situation where the branch I designated for smoking was not smoked; > > another branch was smoked instead. > > > > When I smoke a branch other than blead I simply type the name of the > > non-blead branch into the first record in 'smokecurrent.gitbranch', > > thereby pushing all other records in that file down one line. > > > > At 14:14 UTC I modified 'smokecurrent.gitbranch' to place > > 'smoke-me/khw-locale' in the first position. > > > > ##### > > [p5smoke] $ l smokecurrent.gitbranch > > -rw-r--r-- 1 jkeenan jkeenan 2140 Mar 27 14:14 > > smokecurrent.gitbranch > > > > [p5smoke] $ head -n 5 smokecurrent.gitbranch > > blead > > smoke-me/jkeenan/ntyni/133928-posix-2nd > > blead > > smoke-me/jkeenan/ntyni/133928-posix > > ##### > > > > I then resumed a screen session in which I called 'sh > > ./smokecurrent.sh'. This particular smoke configuration is quite > > thorough, so I did not check on its progress for the next 5 hours. > > When > > I checked just now, I discovered to my surprise that > > 'smoke-me/khw-locale' had not been smoked. Instead, > > 'smoke-me/jkeenan/ntyni/133928-posix-2nd' was smoked -- despite being > > the *third* record in smokecurrent.gitbranch. (I originally smoked > > this > > branch on March 16; see http://perl5.test-smoke.org/report/82582. > > All > > smoke runs between March 16 and March 26 were done on blead.) > > > > I'm attaching a file, > > 'febb9d45.freebsd-11.20190327.smokecurrent.reduced.log', which is > > smokecurrent.log with the data from the individual smoke passes > > stripped > > out. (They were all PASS, so the results are not at issue here.) > > > > In that file, note that, as expected, 'smoke-me/khw-locale' is the > > branch cited at lines 3, 6, 10 and 16, as expected. At line 19, > > 'febb9d45da2878c49f62aae8212fa4d4c13d883c' is cited as the > > 'patchlevel' > > in the 'perl-current/' directory. Commit febb9d45 was HEAD of the > > 'smoke-me/khw-locale' branch at that point in time. This commit > > cited > > again at line 29. > > > > However, at line 30 we suddenly see a different, older branch being > > cited: "Smoking branch smoke-me/jkeenan/ntyni/133928-posix-2nd". > > That > > branch is what gets cited in the web report > > (http://perl5.test-smoke.org/report/83471), even though febb9d45 is > > the > > SHA cited. > > > > Checking on the two repositories which are part of this Test-Smoke > > rig, > > I see that 'git-perl' is in the intended branch but 'perl-current' is > > in > > the non-intended, older branch: > > > > ##### > > [git-perl] $ gitcurr > > smoke-me/khw-locale > > [git-perl] $ cd ../perl-current/ > > [perl-current] $ gitcurr > > 'smoke-me/jkeenan/ntyni/133928-posix-2nd > > ##### > > > > ('gitcurr' is an alias for 'git branch' grepped for the current > > branch > > -- that with the asterisk.) > > > > However, when I call 'git show --format=fuller' in each of these two > > repositories, I get exactly the same output! (See attachments.) > > > > Additional data: The name of the requested branch shows up in the > > *.log > > file for this run. The name of the non-requested branch shows up in > > the > > *.jsn, *.log, *.out and *.rpt files for this run. > > > > ##### > > [smokecurrent] $ grep -l -E 'khw' *febb9d45* > > logfebb9d45da2878c49f62aae8212fa4d4c13d883c.log > > [smokecurrent] $ grep -l -E 'ntyni' *febb9d45* > > jsnfebb9d45da2878c49f62aae8212fa4d4c13d883c.jsn > > logfebb9d45da2878c49f62aae8212fa4d4c13d883c.log > > outfebb9d45da2878c49f62aae8212fa4d4c13d883c.out > > rptfebb9d45da2878c49f62aae8212fa4d4c13d883c.rpt > > ##### > > > > It is possible that the branch I requested to be smoked, > > 'smoke-me/khw-locale', was indeed smoked -- but if so it was > > ultimately > > recorded as being the result of a smoke of the > > 'smoke-me/jkeenan/ntyni/133928-posix-2nd' branch. And it is under > > the > > latter, non-intended name by which this smoke-test result is > > catalogued > > at http://perl5.test-smoke.org/search. > >

> > I had another instance of this problem tonight. I believe that the > branch I intended to smoke (as set in smokecurrent.gitbranch) was > actually smoked, but it was reported as a different branch. See > below. >

> > So, either I've wasted 4+ hours of server time smoking the code of an > > older branch, or I've spent 4+ hours testing the correct branch but > > it's > > not being reported under that branch's name. Which is it? > > > > Can you explain why, starting at line 30 of the log file, the wrong > > branch name is being used? > > > > Thank you very much. > > Jim Keenan

> > Tonight I intended to smoke this branch: smoke-me/khw-petr > > Here is the head of smokecurrent.log. > > ##### > [2019-05-16 22:10:32+0000] ==> Starting synctree > [2019-05-16 22:10:32+0000] qx[/usr/local/bin/git remote update --prune

> 2>&1]

> [2019-05-16 22:10:36+0000] qx[/usr/local/bin/git checkout smoke- > me/khw-petr > [2019-05-16 22:10:36+0000] 2>&1] > Switched to a new branch 'smoke-me/khw-petr' > [2019-05-16 22:10:36+0000] qx[/usr/local/bin/git reset --hard > origin/smoke-me/khw-petr > [2019-05-16 22:10:36+0000] 2>&1] > [2019-05-16 22:10:36+0000] qx[/usr/local/bin/git clean -dfx 2>&1] > [2019-05-16 22:10:37+0000] qx[/usr/local/bin/git fetch origin 2>&1] > [2019-05-16 22:10:39+0000] qx[/usr/local/bin/git checkout smoke- > me/khw-petr > [2019-05-16 22:10:39+0000] 2>&1] > error: Your local changes to the following files would be overwritten > by checkout: > patchlevel.h > Please commit your changes or stash them before you switch branches. > Aborting > [2019-05-16 22:10:39+0000] qx[/usr/local/bin/git reset --hard > origin/smoke-me/khw-petr > [2019-05-16 22:10:39+0000] 2>&1] > [2019-05-16 22:10:39+0000] qx[/usr/local/bin/perl5 > Porting/make_dot_patch.pl > .patch] > [2019-05-16 22:10:39+0000] /home/jkeenan/p5smoke/perl-current is now > up to patchlevel 130031a372ca43b9e664276be0cb4986f39bc432 > [2019-05-16 22:10:39+0000] ==> Starting runsmoke > [2019-05-16 22:10:39+0000] [./tssmokeperl.pl] > chdir(/home/jkeenan/p5smoke/perl-current) > [2019-05-16 22:10:39+0000] Setup alarm: Fri May 17 07:10:39 2019 > "x.log" 209L, 12600C > 1,1 Top > [2019-05-16 22:10:36+0000] qx[/usr/local/bin/git checkout smoke- > me/khw-petr > [2019-05-16 22:10:36+0000] qx[/usr/local/bin/git reset --hard > origin/smoke-me/khw-petr > [2019-05-16 22:10:36+0000] 2>&1] > [2019-05-16 22:10:36+0000] qx[/usr/local/bin/git clean -dfx 2>&1] > [2019-05-16 22:10:39+0000] 2>&1] > Please commit your changes or stash them before you switch branches. > [2019-05-16 22:10:39+0000] 2>&1] > [2019-05-16 22:10:39+0000] ==> Starting runsmoke > [2019-05-16 22:10:39+0000] [./tssmokeperl.pl] > chdir(/home/jkeenan/p5smoke/perl-current) > [2019-05-16 22:10:39+0000] Setup alarm: Fri May 17 07:10:39 2019 > [2019-05-16 22:10:39+0000] qx[/usr/local/bin/perl5 -e "require > q[/home/jkeenan/p5smoke/perl-current/cpan/Test- > Harness/lib/Test/Harness.pm];print Test::Harness->VERSION" 2>&1] > [2019-05-16 22:10:39+0000] Found: Test::Harness version 3.42. > [2019-05-16 22:10:39+0000] Reading build configurations from > /home/jkeenan/p5smoke/perlcurrent.cfg > [2019-05-16 22:10:39+0000] Reading 'Policy.sh' from default content > (v=1) > [2019-05-16 22:10:39+0000] Running test suite only with 'harness' > [2019-05-16 22:10:39+0000] make distclean ... > [2019-05-16 22:10:39+0000] Smoking patch > 130031a372ca43b9e664276be0cb4986f39bc432 v5.30.0-RC1-16-g130031a372 > [2019-05-16 22:10:39+0000] Smoking branch blead > ##### > > Note the last two lines above. 30031a372ca43b9e664276be0cb4986f39bc432 > is indeed the SHA for HEAD in the smoke-me/khw-petr branch. But then > in the last line it states that we're smoking 'blead'. And 'blead' is > what got reported in http://perl5.test-smoke.org/report/87533! > > If I descend into the perl-current directory, I'm told that the > current branch is 'blead'. > > ##### > $ gitcurr > blead > ##### > > But 'git log' starts like this: > > ##### > commit 130031a372ca43b9e664276be0cb4986f39bc432 (HEAD -> blead, > origin/smoke-me/khw-petr) > Author: Karl Williamson <khw@cpan.org> > Date: Thu May 16 15:47:20 2019 -0600 > > [perl #134112] 5.30 regression in Unicode properties defined with > splice > > ##### > > So the right SHA, the wrong branch, but the right "branch based on". > > Can you explain what is happening? > > Thank you very much. > Jim Keenan

And here is the same problem again -- on a completely different machine/OS. See http://perl5.test-smoke.org/report/87530 Note that the SHA reported as having been smoked is 130031a372ca43b9e664276be0cb4986f39bc432 -- the HEAD of the smoke-me/khw-petr branch. But the *branch* reported as having been smoked is: smoke-me/tonyc/133958-realloc-errno-success ... which was the most recent non-blead branch smoked on this machine. ##### $ head -3 smokecurrent.gitbranch smoke-me/khw-petr blead smoke-me/tonyc/133958-realloc-errno-success ##### Once again, smokecurrent.log starts out as indicating that 'smoke-me/khw-petr' is being smoked, and even lists the correct SHA -- but then it says that 'blead' is being smoked. ##### $ head -30 smokecurrent.log [2019-05-16 18:13:08-0400] ==> Starting synctree [2019-05-16 18:13:08-0400] qx[/usr/bin/git remote update --prune 2>&1] [2019-05-16 18:13:13-0400] qx[/usr/bin/git checkout smoke-me/khw-petr [2019-05-16 18:13:13-0400] 2>&1] Switched to a new branch 'smoke-me/khw-petr' [2019-05-16 18:13:13-0400] qx[/usr/bin/git reset --hard origin/smoke-me/khw-petr [2019-05-16 18:13:13-0400] 2>&1] [2019-05-16 18:13:13-0400] qx[/usr/bin/git clean -dfx 2>&1] [2019-05-16 18:13:13-0400] qx[/usr/bin/git fetch origin 2>&1] [2019-05-16 18:13:14-0400] qx[/usr/bin/git checkout smoke-me/khw-petr [2019-05-16 18:13:14-0400] 2>&1] error: Your local changes to the following files would be overwritten by checkout: patchlevel.h Please, commit your changes or stash them before you can switch branches. Aborting [2019-05-16 18:13:14-0400] qx[/usr/bin/git reset --hard origin/smoke-me/khw-petr [2019-05-16 18:13:14-0400] 2>&1] [2019-05-16 18:13:14-0400] qx[/usr/bin/perl Porting/make_dot_patch.pl > .patch] [2019-05-16 18:13:14-0400] /home/jkeenan/p5smoke/perl-current is now up to patchlevel 130031a372ca43b9e664276be0cb4986f39bc432 [2019-05-16 18:13:14-0400] ==> Starting runsmoke [2019-05-16 18:13:14-0400] [./tssmokeperl.pl] chdir(/home/jkeenan/p5smoke/perl-current) [2019-05-16 18:13:14-0400] Setup alarm: Fri May 17 00:13:14 2019 [2019-05-16 18:13:14-0400] qx[/usr/bin/perl -e "require q[/home/jkeenan/p5smoke/perl-current/cpan/Test-Harness/lib/Test/Harness.pm];print Test::Harness->VERSION" 2>&1] [2019-05-16 18:13:14-0400] Found: Test::Harness version 3.42. [2019-05-16 18:13:14-0400] Reading build configurations from /home/jkeenan/p5smoke/perlcurrent.cfg [2019-05-16 18:13:14-0400] Reading 'Policy.sh' from default content (v=1) [2019-05-16 18:13:14-0400] Running test suite only with 'harness' [2019-05-16 18:13:14-0400] make distclean ... [2019-05-16 18:13:14-0400] Smoking patch 130031a372ca43b9e664276be0cb4986f39bc432 v5.30.0-RC1-16-g130031a [2019-05-16 18:13:14-0400] Smoking branch smoke-me/tonyc/133958-realloc-errno-success #####

Sat Jul 06 09:47:48 2019 jkeenan [...] cpan.org - Correspondence added

Message body is not shown because it is too large.

Wed Dec 04 10:22:11 2019 jkeenan [...] cpan.org - Correspondence added

On Sat Jul 06 09:47:48 2019, JKEENAN wrote: [snip] Show quoted text

> > Another instance of the same problem: http://perl5.test- > smoke.org/report/90909 > > Branch requested for smoking: smoke-me/jkeenan/haarg/134265-storable > > SHA at HEAD of branch requested for smoking: > 052bf3ea478f1e981347b34ab49e5a9a1b940707 > > ##### > commit 052bf3ea478f1e981347b34ab49e5a9a1b940707 (HEAD -> 134265- > storable, origin/smoke-me/jkeenan/haarg/134265-storable) > Author: Graham Knop <haarg@haarg.org> > AuthorDate: Sat Apr 20 12:53:06 2019 +0200 > Commit: James E Keenan <jkeenan@cpan.org> > CommitDate: Fri Jul 5 08:03:33 2019 -0400 > > move Storable CAN_FLOCK computation into XS > ##### > > Branch reported in GUI as being tested: blead > > ##### > Automated smoke report for branch blead 5.31.2 patch > 052bf3ea478f1e981347b34ab49e5a9a1b940707 v5.31.1-131-g052bf3ea47 > ##### > > This is from my 'perl-reporter-05' VM which smoke-tests perl on > FreeBSD-13. >

Yet another instance of the "correct SHA smoked but branch reported was incorrect" problem: http://perl5.test-smoke.org/report/100925 Branch requested for smoking: ##### $ head -n 1 smokecurrent.gitbranch smoke-me/jkeenan/lgtm-ae73d7ec2 ##### SHA at HEAD of branch requested for smoking: commit b957d605cb4a2fe83969f60f8b3da323ff4eaeba Branch reported as being tested: smoke-me/jkeenan/niner/gh-14816/niner-blead ##### Automated smoke report for branch smoke-me/jkeenan/niner/gh-14816/niner-blead 5.31.7 patch b957d605cb4a2fe83969f60f8b3da323ff4eaeba v5.31.6-77-gb957d605cb ##### In perl-current directory: ##### $ git log --format=fuller commit b957d605cb4a2fe83969f60f8b3da323ff4eaeba (HEAD -> smoke-me/jkeenan/niner/gh-14816/niner-blead, origin/smoke-me/jkeenan/lgtm-ae73d7ec2) Author: James E Keenan <jkeenan@cpan.org> AuthorDate: Mon Dec 2 21:50:17 2019 -0500 Commit: James E Keenan <jkeenan@cpan.org> CommitDate: Mon Dec 2 22:00:04 2019 -0500 Local variable 'imode' hides a parameter of the same name ##### Reporting consequences: http://perl5.test-smoke.org/report/100925 is reported at: http://perl5.test-smoke.org/search?selected_osnm=openbsd&report_osnm=openbsd&selected_perl=all Arch Host OS Git-id Branch Smoke date Status amd64 perl-reporter-04 openbsd - 6.4 v5.31.6-77-gb957d605cb smoke-me/jkeenan/niner/gh-14816/niner-blead 2019-12-03 23:38:37+00 PASS correct platform, hostname, OS version, Git-id, WRONG branch, correct date, correct results At Tony Cook's http://perl.develop-help.com, branch is reported under the requested branch's name. http://perl.develop-help.com/?b=smoke-me%2Fjkeenan%2Flgtm-ae73d7ec2 Thank you very much. Jim Keenan

Mon May 04 14:31:21 2020 jkeenan [...] cpan.org - Correspondence added

On Wed Dec 04 10:22:11 2019, JKEENAN wrote: Show quoted text

> On Sat Jul 06 09:47:48 2019, JKEENAN wrote: > [snip]

> > > > Another instance of the same problem: http://perl5.test- > > smoke.org/report/90909 > > > > Branch requested for smoking: smoke-me/jkeenan/haarg/134265-storable > > > > SHA at HEAD of branch requested for smoking: > > 052bf3ea478f1e981347b34ab49e5a9a1b940707 > > > > ##### > > commit 052bf3ea478f1e981347b34ab49e5a9a1b940707 (HEAD -> 134265- > > storable, origin/smoke-me/jkeenan/haarg/134265-storable) > > Author: Graham Knop <haarg@haarg.org> > > AuthorDate: Sat Apr 20 12:53:06 2019 +0200 > > Commit: James E Keenan <jkeenan@cpan.org> > > CommitDate: Fri Jul 5 08:03:33 2019 -0400 > > > > move Storable CAN_FLOCK computation into XS > > ##### > > > > Branch reported in GUI as being tested: blead > > > > ##### > > Automated smoke report for branch blead 5.31.2 patch > > 052bf3ea478f1e981347b34ab49e5a9a1b940707 v5.31.1-131-g052bf3ea47 > > ##### > > > > This is from my 'perl-reporter-05' VM which smoke-tests perl on > > FreeBSD-13. > >

> > Yet another instance of the "correct SHA smoked but branch reported > was incorrect" problem: > http://perl5.test-smoke.org/report/100925 > > Branch requested for smoking: > ##### > $ head -n 1 smokecurrent.gitbranch > smoke-me/jkeenan/lgtm-ae73d7ec2 > ##### > > SHA at HEAD of branch requested for smoking: > commit b957d605cb4a2fe83969f60f8b3da323ff4eaeba > > Branch reported as being tested: > smoke-me/jkeenan/niner/gh-14816/niner-blead > ##### > Automated smoke report for branch smoke-me/jkeenan/niner/gh- > 14816/niner-blead 5.31.7 patch > b957d605cb4a2fe83969f60f8b3da323ff4eaeba v5.31.6-77-gb957d605cb > ##### > > In perl-current directory: > ##### > $ git log --format=fuller > commit b957d605cb4a2fe83969f60f8b3da323ff4eaeba (HEAD -> smoke- > me/jkeenan/niner/gh-14816/niner-blead, origin/smoke-me/jkeenan/lgtm- > ae73d7ec2) > Author: James E Keenan <jkeenan@cpan.org> > AuthorDate: Mon Dec 2 21:50:17 2019 -0500 > Commit: James E Keenan <jkeenan@cpan.org> > CommitDate: Mon Dec 2 22:00:04 2019 -0500 > > Local variable 'imode' hides a parameter of the same name > ##### > > Reporting consequences: > > http://perl5.test-smoke.org/report/100925 is reported at: > http://perl5.test- > smoke.org/search?selected_osnm=openbsd&report_osnm=openbsd&selected_perl=all > > Arch Host OS Git-id Branch Smoke date Status > amd64 perl-reporter-04 openbsd - 6.4 v5.31.6-77-gb957d605cb > smoke-me/jkeenan/niner/gh-14816/niner-blead 2019-12-03 23:38:37+00 > PASS > > correct platform, hostname, OS version, Git-id, WRONG branch, correct > date, correct results > > At Tony Cook's http://perl.develop-help.com, branch is reported under > the requested branch's name. > http://perl.develop-help.com/?b=smoke-me%2Fjkeenan%2Flgtm-ae73d7ec2 > > Thank you very much. > Jim Keenan

And still another instance of a branch being smoked as intended but reported incorrectly: ##### http://perl5.test-smoke.org/report/112544 ##### Above URL starts with: ##### Automated smoke report for branch blead 5.31.11 patch 8fc4f28f17d7062b4400738f796902b81d39a70c v5.31.10-74-g8fc4f28f17 ##### But SHA 8fc4f28f17d is from the smoke-me/tonyc/17268-yyl_try-recursion-test branch: ##### commit 8fc4f28f17d7062b4400738f796902b81d39a70c (HEAD -> smoke-me/tonyc/17268-yyl_try-recursion-test, origin/smoke-me/tonyc/17268-yyl_try-recursion-test) Author: Tony Cook <tony@develop-help.com> AuthorDate: Mon Apr 6 15:33:08 2020 +1000 Commit: Tony Cook <tony@develop-help.com> CommitDate: Mon Apr 27 11:35:48 2020 +1000 ##### smoke-me/tonyc/17268-yyl_try-recursion-test was the branch I requested: ##### $ head -2 smokecurrent.gitbranch smoke-me/tonyc/17268-yyl_try-recursion-test blead ##### And this report was correctly filed at http://perl.develop-help.com/?b=smoke-me%2Ftonyc%2F17268-yyl_try-recursion-test. Curiously, the same commit was both correctly run and reported in an OpenBSD-6.6 VM which sits on the host on which the above smoke-test was run. ##### http://perl5.test-smoke.org/report/112546 ##### Thank you very much. Jim Keenan

Tue May 19 07:24:35 2020 abeltje [...] cpan.org - Correspondence added

After looking at this again, I found a problem with the code: The 'gitbranchfile' mechanism was designed to have 1 line with a single branch-name in it (without \n), you put in multiple lines with different branches. For now, I'm not going to change the 'single branch name' policy and will only smoke the branch that is on the first line of the file, but I've fixed the newline issue. * I will put this in dev-release 1.77_02 The multi branch smoke is quite hard to solve from within Test::Smoke as it was never designed to work that way, you may have to solve that at a higher level with a shell-script. Thanks for reporting this issue + good luck. an behalf of the TestSmokeTeam, -- Abe.

Tue May 19 07:24:36 2020 abeltje [...] cpan.org - Status changed from 'open' to 'stalled'

Tue May 19 08:00:20 2020 jkeenan [...] cpan.org - Correspondence added

On Tue May 19 07:24:35 2020, ABELTJE wrote: Show quoted text

> After looking at this again, I found a problem with the code: > > The 'gitbranchfile' mechanism was designed to have 1 line with a > single branch-name in it (without \n), you put in multiple lines with > different branches. > For now, I'm not going to change the 'single branch name' policy and > will only smoke the branch that is on the first line of the file, but > I've fixed the newline issue.

In version 0.029 of lib/Test/Smoke/Syncer/Git.pm, we have this documentation and code: ##### =head2 $git->get_git_branch() Reads the first line of the file set in B<gitbranchfile> and returns its value. =cut sub get_git_branch { my $self = shift; return $self->{gitdfbranch} if !$self->{gitbranchfile}; return $self->{gitdfbranch} if ! -f $self->{gitbranchfile}; if (open my $fh, '<', $self->{gitbranchfile}) { $self->log_debug("Reading branch to smoke from: '$self->{gitbranchfile}'"); my $branch = <$fh>; close $fh; return $branch; } $self->log_warn("Error opening '$self->{gitbranchfile}': $!"); return $self->{gitdfbranch}; } ##### I've always interpeted the documentation to mean: "Use the branch cited on the *first* line of gitbranchfile; all other lines in that file will be ignored for the purpose of this smoke-test run." For someone who tests a variety of branches and kicks off smoke-test runs manually, not (evidently) being required to keep the file at just one line has generally been very convenient. $branch should be chomp-ed. Even if a file has just one record in it, my default assumption is that that file ends in a newline (even if that's not strictly true in Unix). Show quoted text

> > * I will put this in dev-release 1.77_02 > > The multi branch smoke is quite hard to solve from within Test::Smoke > as it was never designed to work that way, you may have to solve that > at a higher level with a shell-script. >

I'm not sure what you mean by the "multi branch smoke". I only smoke one branch at a time. If, say, I did want to smoke 3 specific branches on a given day, I would handle that at the outermost level of the Test-Smoke world such as 'smokecurrent.sh'. Show quoted text

> Thanks for reporting this issue + > good luck. > > an behalf of the TestSmokeTeam, > > -- Abe.

Tue May 19 08:00:21 2020 The RT System itself - Status changed from 'stalled' to 'open'