Re: Strange "make check" error
- Original Message - From: "David Kastrup" To: Sent: Friday, May 01, 2015 9:47 AM Subject: Strange "make check" error Never seen anything like that. I suppose it will be a fluke and rerunning the test will "fix" it, but has anybody encountered anything like that? Strange bugout in the middle of the image comparisons? -- David Kastrup Not seen it myself. Possibly a race condition or somesuch? Interesting that it was on output-distance, which is pretty much guaranteed to show a difference. -- Phil Holmes ___ lilypond-devel mailing list lilypond-devel@gnu.org https://lists.gnu.org/mailman/listinfo/lilypond-devel
Re: Strange "make check" error
"Phil Holmes" writes: > - Original Message - > From: "David Kastrup" > To: > Sent: Friday, May 01, 2015 9:47 AM > Subject: Strange "make check" error > > >> >> Never seen anything like that. I suppose it will be a fluke and >> rerunning the test will "fix" it, but has anybody encountered anything >> like that? Strange bugout in the middle of the image comparisons? >> >> -- >> David Kastrup > > > Not seen it myself. Possibly a race condition or somesuch? > > Interesting that it was on output-distance, which is pretty much > guaranteed to show a difference. Oh superb. Reproducible. invoking gs -sDEVICE=png16m -dGraphicsAlphaBits=4 -dTextAlphaBits=4 -slilypond-datadir=input/regression/out-test/share/lilypond/current -r101 -sOutputFile=/tmp/build-lilypond-autobuild/out/test-results/input/regression/out-test/test-output-distance.png -dNOSAFER -dEPSCrop -q -dNOPAUSE input/regression/out-test/test-output-distance.eps -c quit pipe file /tmp/build-lilypond-autobuild/out/test-results/input/regression/out-test-baseline/test-output-distance.png pipe file /tmp/build-lilypond-autobuild/out/test-results/input/regression/out-test/test-output-distance.png dir is /tmp/tmpGUycW1 invoking convert -depth 8 -crop 216x107+0+0 /tmp/build-lilypond-autobuild/out/test-results/input/regression/out-test-baseline/test-output-distance.png /tmp/tmpGUycW1/crop1.png invoking convert -depth 8 -crop 216x107+0+0 /tmp/build-lilypond-autobuild/out/test-results/input/regression/out-test/test-output-distance.png /tmp/tmpGUycW1/crop2.png invoking compare -depth 8 -dissimilarity-threshold 1 /tmp/tmpGUycW1/crop1.png /tmp/tmpGUycW1/crop2.png /tmp/tmpGUycW1/diff.png Traceback (most recent call last): File "/tmp/build-lilypond-autobuild/scripts/build/out/output-distance", line 1342, in main () File "/tmp/build-lilypond-autobuild/scripts/build/out/output-distance", line 1339, in main compare_tree_pairs (zip (args[0::2], args[1::2]), out, options.threshold) File "/tmp/build-lilypond-autobuild/scripts/build/out/output-distance", line 1060, in compare_tree_pairs data.create_html_result_page (dest_dir, threshold) File "/tmp/build-lilypond-autobuild/scripts/build/out/output-distance", line 1042, in create_html_result_page link.link_files_for_html (dest_dir) File "/tmp/build-lilypond-autobuild/scripts/build/out/output-distance", line 674, in link_files_for_html compare_png_images (old, new, dest_dir) File "/tmp/build-lilypond-autobuild/scripts/build/out/output-distance", line 92, in compare_png_images system ('compare -depth 8 -dissimilarity-threshold 1 %(dir)s/crop1.png %(dir)s/crop2.png %(dir)s/diff.png' % locals ()) File "/tmp/build-lilypond-autobuild/scripts/build/out/output-distance", line 1090, in system assert stat == 0 AssertionError rm -rf /tmp/tmpGUycW1 /tmp/lilypond-autobuild/GNUmakefile.in:325: recipe for target 'local-check' failed And it's right at the start of some phase. I suspect that something with my recent update to Ubuntu 15.04 is at play here. Sigh. -- David Kastrup ___ lilypond-devel mailing list lilypond-devel@gnu.org https://lists.gnu.org/mailman/listinfo/lilypond-devel
Re: Strange "make check" error
David Kastrup writes: > "Phil Holmes" writes: > >> - Original Message - >> From: "David Kastrup" >> To: >> Sent: Friday, May 01, 2015 9:47 AM >> Subject: Strange "make check" error >> >> >>> >>> Never seen anything like that. I suppose it will be a fluke and >>> rerunning the test will "fix" it, but has anybody encountered anything >>> like that? Strange bugout in the middle of the image comparisons? >>> >>> -- >>> David Kastrup >> >> >> Not seen it myself. Possibly a race condition or somesuch? >> >> Interesting that it was on output-distance, which is pretty much >> guaranteed to show a difference. > > Oh superb. Reproducible. Well, could also be that my baseline is broken. So need to redo from full scratch. -- David Kastrup ___ lilypond-devel mailing list lilypond-devel@gnu.org https://lists.gnu.org/mailman/listinfo/lilypond-devel
Re: Strange "make check" error
On 01/05/15 10:28, David Kastrup wrote: David Kastrup writes: "Phil Holmes" writes: - Original Message - From: "David Kastrup" To: Sent: Friday, May 01, 2015 9:47 AM Subject: Strange "make check" error Never seen anything like that. I suppose it will be a fluke and rerunning the test will "fix" it, but has anybody encountered anything like that? Strange bugout in the middle of the image comparisons? -- David Kastrup Not seen it myself. Possibly a race condition or somesuch? Interesting that it was on output-distance, which is pretty much guaranteed to show a difference. Oh superb. Reproducible. Well, could also be that my baseline is broken. So need to redo from full scratch. Are you using Patchy and have you been using Patchy for different patches recently? I sometimes get odd assertions all the time when running a number of different patches one after another, even though patchy is supposed to use make clean (or whatever it does) after each test has completed. I can usually smell when something isn't right or looks odd and find that the baseline or even the initial test-master / test-staging branches that patchy uses are 'broken'; so I git branch -D both those test-* branches, cat /dev/null >~/.lilypond-patchy-cache (which removes any cruft that is probably overkill) and rm -rf /tmp/*lily* in case the place that patchy puts the test branches to run the test are poisoned. and then re-run ./test-patches.py and that usually helps. James ___ lilypond-devel mailing list lilypond-devel@gnu.org https://lists.gnu.org/mailman/listinfo/lilypond-devel
Re: Strange "make check" error
James writes: > On 01/05/15 10:28, David Kastrup wrote: >> David Kastrup writes: >> >>> "Phil Holmes" writes: >>> - Original Message - From: "David Kastrup" To: Sent: Friday, May 01, 2015 9:47 AM Subject: Strange "make check" error > Never seen anything like that. I suppose it will be a fluke and > rerunning the test will "fix" it, but has anybody encountered anything > like that? Strange bugout in the middle of the image comparisons? > > -- > David Kastrup Not seen it myself. Possibly a race condition or somesuch? Interesting that it was on output-distance, which is pretty much guaranteed to show a difference. >>> Oh superb. Reproducible. >> Well, could also be that my baseline is broken. So need to redo from >> full scratch. >> > Are you using Patchy and have you been using Patchy for different > patches recently? I sometimes get odd assertions all the time when > running a number of different patches one after another, even though > patchy is supposed to use make clean (or whatever it does) after each > test has completed. > > I can usually smell when something isn't right or looks odd and find > that the baseline or even the initial test-master / test-staging > branches that patchy uses are 'broken'; so I git branch -D both those > test-* branches, cat /dev/null >~/.lilypond-patchy-cache (which > removes any cruft that is probably overkill) and rm -rf /tmp/*lily* in > case the place that patchy puts the test branches to run the test are > poisoned. > > and then re-run ./test-patches.py and that usually helps. Cleaned everything out. Baseline got rebuilt, same error. Checking different patch (4360 instead of 4357), same error. I'll eventually do the "reboot computer" dance but I suspect that we got ourselves a problem with Ubuntu 15.04. -- David Kastrup ___ lilypond-devel mailing list lilypond-devel@gnu.org https://lists.gnu.org/mailman/listinfo/lilypond-devel
Re: Strange "make check" error
David Kastrup writes: > Cleaned everything out. Baseline got rebuilt, same error. Checking > different patch (4360 instead of 4357), same error. > > I'll eventually do the "reboot computer" dance but I suspect that we got > ourselves a problem with Ubuntu 15.04. Well, either that, or with the latest changes in master. But they should have passed through some instance of Patchy too after all. Or doesn't the staging patchy do a "make check"? Perhaps it only does "make test"? Because it is "make check" that appears to be newly broken. -- David Kastrup ___ lilypond-devel mailing list lilypond-devel@gnu.org https://lists.gnu.org/mailman/listinfo/lilypond-devel
Re: Strange "make check" error
On 01/05/15 11:40, David Kastrup wrote: David Kastrup writes: Cleaned everything out. Baseline got rebuilt, same error. Checking different patch (4360 instead of 4357), same error. I'll eventually do the "reboot computer" dance but I suspect that we got ourselves a problem with Ubuntu 15.04. Well, either that, or with the latest changes in master. But they should have passed through some instance of Patchy too after all. Or doesn't the staging patchy do a "make check"? Perhaps it only does "make test"? Because it is "make check" that appears to be newly broken. Patchy staging doesn't do the make check but it does do all the others including make doc --snip-- From ssh://git.sv.gnu.org/srv/git/lilypond 8dd74d9..e108145 staging-> origin/staging Merged staging, now at:e1081454de4c8167f3f010912de06186f61d Success:./autogen.sh --noconfigure Success:/tmp/lilypond-autobuild/configure --disable-optimising Success:nice make clean Success:nice make -j6 CPU_COUNT=6 Success:nice make test -j6 CPU_COUNT=6 Success:nice make doc -j6 CPU_COUNT=6 To ssh://pkx1...@git.sv.gnu.org/srv/git/lilypond.git 8dd74d9..e108145 test-staging -> master Success:pushed to master --snip-- ___ lilypond-devel mailing list lilypond-devel@gnu.org https://lists.gnu.org/mailman/listinfo/lilypond-devel
Re: Strange "make check" error
On 01/05/15 11:36, David Kastrup wrote: James writes: On 01/05/15 10:28, David Kastrup wrote: David Kastrup writes: "Phil Holmes" writes: - Original Message - From: "David Kastrup" To: Sent: Friday, May 01, 2015 9:47 AM Subject: Strange "make check" error Never seen anything like that. I suppose it will be a fluke and rerunning the test will "fix" it, but has anybody encountered anything like that? Strange bugout in the middle of the image comparisons? -- David Kastrup Not seen it myself. Possibly a race condition or somesuch? Interesting that it was on output-distance, which is pretty much guaranteed to show a difference. Oh superb. Reproducible. Well, could also be that my baseline is broken. So need to redo from full scratch. Are you using Patchy and have you been using Patchy for different patches recently? I sometimes get odd assertions all the time when running a number of different patches one after another, even though patchy is supposed to use make clean (or whatever it does) after each test has completed. I can usually smell when something isn't right or looks odd and find that the baseline or even the initial test-master / test-staging branches that patchy uses are 'broken'; so I git branch -D both those test-* branches, cat /dev/null >~/.lilypond-patchy-cache (which removes any cruft that is probably overkill) and rm -rf /tmp/*lily* in case the place that patchy puts the test branches to run the test are poisoned. and then re-run ./test-patches.py and that usually helps. Cleaned everything out. Baseline got rebuilt, same error. Checking different patch (4360 instead of 4357), same error. I'll eventually do the "reboot computer" dance but I suspect that we got ourselves a problem with Ubuntu 15.04. Could be; I am using Mint 17.1 which is still based on 14:04. https://en.wikipedia.org/wiki/List_of_Linux_Mint_releases James ___ lilypond-devel mailing list lilypond-devel@gnu.org https://lists.gnu.org/mailman/listinfo/lilypond-devel
Re: Strange "make check" error
> I suspect that we got ourselves a problem with Ubuntu 15.04. I got the same assertion, on Lubuntu 15.04. p ___ lilypond-devel mailing list lilypond-devel@gnu.org https://lists.gnu.org/mailman/listinfo/lilypond-devel
Re: Strange "make check" error
David Kastrup writes: > David Kastrup writes: > >> Cleaned everything out. Baseline got rebuilt, same error. Checking >> different patch (4360 instead of 4357), same error. >> >> I'll eventually do the "reboot computer" dance but I suspect that we got >> ourselves a problem with Ubuntu 15.04. > > Well, either that, or with the latest changes in master. But they > should have passed through some instance of Patchy too after all. Or > doesn't the staging patchy do a "make check"? Perhaps it only does > "make test"? Because it is "make check" that appears to be newly > broken. Oh good grief. man compare The compare program returns 2 on error otherwise 0 if the images are similar or 1 if they are dissimilar. Line 1085 in scripts/build/output-distance.py: def system (x): print 'invoking', x stat = os.system (x) assert stat == 0 Now "compare" is called with the option -dissimilarity-threshold 1 which presumably should cater for an exit status of 0. I cannot manage with _any_ options to get a status of 0 even for completely identical images. Probably a bug in this version of compare. -- David Kastrup ___ lilypond-devel mailing list lilypond-devel@gnu.org https://lists.gnu.org/mailman/listinfo/lilypond-devel