Bug 194930
| Summary: | make index fails with "make_index: Circular dependency loop found: pulseaudio-0.9.23_3 depends upon itself." | ||
|---|---|---|---|
| Product: | Ports & Packages | Reporter: | Thomas Schweikle <tps> |
| Component: | Ports Framework | Assignee: | Port Management Team <portmgr> |
| Status: | Closed Unable to Reproduce | ||
| Severity: | Affects Only Me | CC: | swills |
| Priority: | --- | ||
| Version: | Latest | ||
| Hardware: | Any | ||
| OS: | Any | ||
|
Description
Thomas Schweikle
2014-11-10 18:00:56 UTC
still valid? I would think there would break portsnap if it were happening universally. silence is golden... Wasn't available for some time -- training course ... It is the same again: --- describe.x11-toolkits --- --- describe.x11-wm --- make_index: Circular dependency loop found: pulseaudio-0.9.23_3 depends upon itself. What I did: # cd /usr/ports # rm -rf * .arcconfig .gitignore .svn # svn co https://svn0.eu.freebsd.org/ports/head /usr/ports Since this leads to an 1:1 copy if all would be OK theere should be no message like that ..., but: # make index leads to the above. Maybe a mirror problem with svn0.eu.freebsd.org? (In reply to John Marino from comment #1) > I would think there would break portsnap if it were happening universally. You didn't address why we aren't seeing fallout across the board. Doesn't everything depend on index? I don't even know who I would assign this PR to. my guess is you have something in make.conf that is affecting dependencies. Something like make index should be done in clean chroot. Just a guess though. To verify my assumtion it could be a mirror problem I did: # cd /usr/ports # rm -rf * .arcconfig .gitignore .svn # svn co https://svn0.us-east.freebsd.org/ports/head /usr/ports # make index ... --- describe.x11-toolkits --- --- describe.x11-wm --- make_index: Circular dependency loop found: jackit-0.124.1 depends upon itself. :-( Next test: # cd /usr/ports # rm -rf * .arcconfig .gitignore .svn # svn co https://svn0.us-west.freebsd.org/ports/head /usr/ports # make index ... --- describe.x11-toolkits --- --- describe.x11-wm --- make_index: Circular dependency loop found: jackit-0.124.1 depends upon itself. :-( Thus again back to svn0.eu.freebsd.org: # cd /usr/ports # rm -rf * .arcconfig .gitignore .svn # svn co https://svn0.eu.freebsd.org/ports/head /usr/ports # make index ... --- describe.x11-toolkits --- --- describe.x11-wm --- make_index: Circular dependency loop found: jackit-0.124.1 depends upon itself. :-( Again. Since your guess: # cat /etc/make.conf DISABLE_SIZE=yes #MAKE_JOBS_NUMBER=5 # PKGNG WITH_PKGNG="YES" # SUBTHREADS=2 OPTIONS_UNSET=X11 BATCH=yo # PERL_VERSION=5.16.3 # DEFAULT_VERSIONS=pgsql=9.3 --- eof But this leads to an other suggestion: the machine running "make index" on has four processors. Since MAKE_JOBS_NUMBER is unset I could try to set it to one. Did it in /etc/make.conf, then had a look at what is going on: # ps axd ... 887 1 Is 0:00.15 | |-- /bin/csh 60899 1 I+ 0:00.02 | | `-- make index 60954 1 I+ 0:00.02 | | `-- make /usr/ports/INDEX-10 60957 1 I+ 0:00.00 | | `-- /bin/sh -e -c echo -n "Generating INDEX-10 - please wait.."; if [ "" != "" ]; then exp 60959 1 I+ 0:00.00 | | `-- /bin/sh -e -c echo -n "Generating INDEX-10 - please wait.."; if [ "" != "" ]; then e 60960 1 S+ 0:00.03 | | `-- make -j4 INDEX_TMPDIR=/tmp/index.x0rfIgkD BUILDING_INDEX=1 __MAKE_SHELL=/rescue/sh E 70208 1 I 0:00.00 | | |-- sh -ev 70211 1 I 0:00.02 | | | `-- make -B _databases.describe 70226 1 I 0:00.16 | | | `-- make -B describe DIRPRFX=databases/ 70297 1 S 0:00.14 | | | `-- /rescue/sh -e -c for sub in R-cran-DBI R-cran-RMySQL R-cran-RSQLite R-cran-R 92800 1 S 0:00.01 | | | `-- make -B describe 92817 1 RL 0:00.01 | | | `-- make -V PORTVERSION -C /usr/ports/lang/python27 72828 1 I 0:00.00 | | |-- sh -ev 72829 1 I 0:00.02 | | | `-- make -B _devel.describe 72843 1 I 0:00.84 | | | `-- make -B describe DIRPRFX=devel/ 73219 1 S 0:00.12 | | | `-- /rescue/sh -e -c for sub in 9base ChipmunkPhysics ElectricFence ORBit2 ORBit 92811 1 S 0:00.00 | | | `-- make -B describe 92818 1 R 0:00.00 | | | `-- /usr/local/sbin/pkg-static -v 86295 1 S 0:00.00 | | |-- sh -ev 86297 1 S 0:00.02 | | | `-- make -B _editors.describe 86313 1 S 0:00.05 | | | `-- make -B describe DIRPRFX=editors/ 86336 1 S 0:00.04 | | | `-- /rescue/sh -e -c for sub in 2bsd-vi abiword abiword-docs aee aewan apel ased 92799 1 R 0:00.01 | | | `-- make -B describe 88797 1 S 0:00.00 | | `-- sh -ev 88799 1 S 0:00.02 | | `-- make -B _emulators.describe 88813 1 S 0:00.03 | | `-- make -B describe DIRPRFX=emulators/ 88833 1 S 0:00.02 | | `-- /rescue/sh -e -c for sub in adamem advancemame advancemenu advancemess aftp 92805 1 S 0:00.00 | | `-- make -B describe 92816 1 R 0:00.00 | | `-- /usr/local/sbin/pkg-static version -t 1.3.8 1.3.8 ... This where the commands given: # cd /usr/ports # make index ... --- describe.x11-toolkits --- --- describe.x11-wm --- make_index: Circular dependency loop found: jackit-0.124.1 depends upon itself. Interestingly "make" ignores what I told it in "/etc/make.conf". who do you want me to assign the PR to? Obviously it's not a global issue. Since all tests on the quad-core system failed, I decided to have a look at a single core machine. Same setup, same OS, but only one core: # cd /usrPorts # rm -rf * .arcconfig .gitignore .svn # svn co https://svn0.eu.freebsd.org/ports/head /usr/ports # make index ... # :-) this works!! Conclusion: "make index" has, since changes made to use multiple cores, problems generating the index. Can anyone confirm? Since it is something that has to do with general ports setup and maintenance, I'd say someone who has to do with ports setup and maintenance. Some additional tests give a really clear result: if "make index" is called with option "-jX" with "X" set to anything else than "1" it will fail. An other try, this time with a single core system, but "-j4" (extraordinary slow) gave the awaited result: circular dependency error and no index generated. try on your quad with DISABLE_MAKE_JOBS=yes set And I wanted an actual name tbh. I can add portmgr to the list but that will have varied results. (In reply to John Marino from comment #12) > try on your quad with DISABLE_MAKE_JOBS=yes set > > And I wanted an actual name tbh. I can add portmgr to the list but that > will have varied results. actually your make file commented out make jobs, your quad should be -j1 anyway. (In reply to Thomas Schweikle from comment #8) > But this leads to an other suggestion: the machine running "make index" on > has four processors. Since MAKE_JOBS_NUMBER is unset I could try to set it > to one. oh, you already experimented. make index uses multiple processors no matter what. the variable is INDEX_JOBS e.g. make index INDEX_JOBS=1 Bryan was the last person to have a non-prop commit. I'll assign the PR to him. Bryan, if this is wrong, just reassign it however you think is best. Tried additional tests. Set "INDEX_JOBS=1", more than one core -> Fails: # make index Generating INDEX-10 - please wait.. make_index: Circular dependency loop found: jackit-0.124.1 depends upon itself. # Set "INDEX_JOBS=1", one core -> Fails: # make index Generating INDEX-10 - please wait.. make_index: Circular dependency loop found: jackit-0.124.1 depends upon itself. # Set "INDEX_JOBS=4", one core -> Fails: # make index Generating INDEX-10 - please wait.. make_index: Circular dependency loop found: jackit-0.124.1 depends upon itself. # all at revision 372810! As this is not really what I awaited for I'll do another view tests. This would be a portsnap bug (if even a bug). Note I have not analyzed this at all. Assigning to portmgr@ for now. Tried building index for port revisions 372804 372817 372819 372821 372824 372825 372826 for all it was the same: make_index: Circular dependency loop found: jackit-0.124.1 depends upon itself. I've checked out the whole ports tree into a clean directory every time. This on five systems all at the same level: FreeBSD test-sct-bf-muc 10.1-STABLE FreeBSD 10.1-STABLE #93 r274683: Wed Nov 19 06:02:56 CET 2014 root@test-sct-bf-muc:/usr/obj/usr/src/sys/TEST-SCT-BF-MUC amd64 Can't reproduce this issue and it's a bit old. Re-open if there's still a problem. Thanks! |
