Your message dated Mon, 11 Apr 2016 21:38:58 -0700 with message-id <[email protected]> and subject line Re: Bug#820658: make: filesystem timestamp race: outdated target does not get updated if a missing dependency gets created with the same timestamp has caused the Debian Bug report #820658, regarding make: filesystem timestamp race: outdated target does not get updated if a missing dependency gets created with the same timestamp to be marked as done.
This means that you claim that the problem has been dealt with. If this is not the case it is now your responsibility to reopen the Bug report if necessary, and/or fix the problem forthwith. (NB: If you are a system administrator and have no idea what this message is talking about, this may indicate a serious mail system misconfiguration somewhere. Please contact [email protected] immediately.) -- 820658: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=820658 Debian Bug Tracking System Contact [email protected] with problems
--- Begin Message ---Package: make Version: 3.81-8.2 Severity: grave Justification: renders package unusable Control: block 725629 with -1 Hi, consider this Makefile: ----- Makefile/Makefile.in ----- all: do_something foo: echo '# foo' > $@ Makefile: Makefile.in foo cat Makfile.in foo > $@ ----- ----- If 'foo' does not exist, any execution of the 'all' target should first create 'foo' and recreate 'Makefile'. This works fine, unless foo is created with the same timestamp as the existing Makefile. Attached is a slightly more advanced proof-of-concept Makefile.in, run it as time make -f Makefile.in loop It should run forever, but on all machines where I tried it, it terminated after some time because Makefile was not updated: $ time make -f Makefile.in loop make clean make[1]: Entering directory '/home/andreas/725629' rm -f blah cp Makefile.in Makefile make[1]: Leaving directory '/home/andreas/725629' set -x ; while make test ; do make clean ; done + make test make[1]: Entering directory '/home/andreas/725629' echo 'foobar:' >blah echo ' @echo FOOBAR' >>blah cat Makefile.in blah > Makefile if ! grep -q ^foobar: Makefile ; then echo FAIL ; ls -lart --time-style=full-iso Makefile* blah ; exit 1 ; else echo OK ; fi OK make[1]: Leaving directory '/home/andreas/725629' + make clean make[1]: Entering directory '/home/andreas/725629' rm -f blah cp Makefile.in Makefile make[1]: Leaving directory '/home/andreas/725629' [... skipped some iterations ...] + make test make[1]: Entering directory '/home/andreas/725629' echo 'foobar:' >blah echo ' @echo FOOBAR' >>blah cat Makefile.in blah > Makefile if ! grep -q ^foobar: Makefile ; then echo FAIL ; ls -lart --time-style=full-iso Makefile* blah ; exit 1 ; else echo OK ; fi OK make[1]: Leaving directory '/home/andreas/725629' + make clean make[1]: Entering directory '/home/andreas/725629' rm -f blah cp Makefile.in Makefile make[1]: Leaving directory '/home/andreas/725629' + make test make[1]: Entering directory '/home/andreas/725629' echo 'foobar:' >blah echo ' @echo FOOBAR' >>blah if ! grep -q ^foobar: Makefile ; then echo FAIL ; ls -lart --time-style=full-iso Makefile* blah ; exit 1 ; else echo OK ; fi FAIL -rw-r--r-- 1 andreas andreas 377 2016-04-11 03:03:37.588201637 +0200 Makefile.in -rw-r--r-- 1 andreas andreas 22 2016-04-11 03:40:50.757214074 +0200 blah -rw-r--r-- 1 andreas andreas 377 2016-04-11 03:40:50.757214074 +0200 Makefile Makefile:2: recipe for target 'test' failed make[1]: *** [test] Error 1 make[1]: Leaving directory '/home/andreas/725629' real 0m0.152s user 0m0.012s sys 0m0.016s It terminates even faster on a filesystem with one-second-resolution: anbe@exodar:~/725629$ time make -f Makefile.in loop make clean make[1]: Entering directory '/home/anbe/725629' rm -f blah cp Makefile.in Makefile make[1]: Leaving directory '/home/anbe/725629' set -x ; while make test ; do make clean ; done + make test make[1]: Entering directory '/home/anbe/725629' echo 'foobar:' >blah echo ' @echo FOOBAR' >>blah if ! grep -q ^foobar: Makefile ; then echo FAIL ; ls -lart --time-style=full-iso Makefile* blah ; exit 1 ; else echo OK ; fi FAIL -rw-r--r-- 1 anbe anbe 266 2016-04-11 02:45:46.000000000 +0200 Makefile.in.old -rw-r--r-- 1 anbe anbe 377 2016-04-11 03:08:04.000000000 +0200 Makefile.in -rw-r--r-- 1 anbe anbe 22 2016-04-11 03:09:25.000000000 +0200 blah -rw-r--r-- 1 anbe anbe 377 2016-04-11 03:09:25.000000000 +0200 Makefile Makefile:2: recipe for target 'test' failed make[1]: *** [test] Error 1 make[1]: Leaving directory '/home/anbe/725629' real 0m0.080s user 0m0.000s sys 0m0.000s I could reproduce this on different architectures, with different filesystems (ppc64el/ext4, hurd-i386/ext2 amd64/xfs,tmpfs) and with the make packages from wheezy, jessie and stretch. An example where this bug (non-deterministically) breaks the build process on some architectures, is #725629. Andreastest: Makefile if ! grep -q ^foobar: Makefile ; then echo FAIL ; ls -lart --time-style=full-iso Makefile* blah ; exit 1 ; else echo OK ; fi Makefile: Makefile.in blah cat Makefile.in blah > $@ blah: echo 'foobar:' >$@ echo ' @echo FOOBAR' >>$@ clean: rm -f blah cp Makefile.in Makefile loop: Makefile make clean set -x ; while make test ; do make clean ; done
--- End Message ---
--- Begin Message ---On Mon, Apr 11 2016, Andreas Beckmann wrote: > Package: make > Version: 3.81-8.2 > Severity: grave > Justification: renders package unusable > Control: block 725629 with -1 > Hi, > > consider this Makefile: > > ----- Makefile/Makefile.in ----- > all: > do_something > foo: > echo '# foo' > $@ > Makefile: Makefile.in foo > cat Makfile.in foo > $@ > ----- ----- > > If 'foo' does not exist, any execution of the 'all' target should first > create 'foo' and recreate 'Makefile'. This works fine, unless foo is > created with the same timestamp as the existing Makefile. Yes, you are running into a race condition. But this is documented make behaviour: Next: Reading, Prev: Overview, Up: Overview Preparing and Running Make ========================== The 'make' program uses the makefile data base and the last-modification times of the files to decide which of the files need to be updated. For each of those files, it issues the recipes recorded in the data base. If the timestamps are the same, then the target foo is not newer than the target Makefile, and is thus not remade. Looking the other way, Makefile is not out of date compared to foo, so it does not need to be refreshed. You could easily add a delay if you want to avoid the race condition. manoj -- Opportunities are usually disguised as hard work, so most people don't recognize them. Manoj Srivastava <[email protected]> <http://www.golden-gryphon.com/> 4096R/C5779A1C E37E 5EC5 2A01 DA25 AD20 05B6 CF48 9438 C577 9A1C
smime.p7s
Description: S/MIME cryptographic signature
--- End Message ---
