Quick recommendation, if you build bison before flex and then, build flex after, delete the bison source tree and restart the build from scratch. I have found out that the configure script for bison need to pick up flex for the testsuite to run.
Alain 2017-08-17 22:15 GMT-04:00 Ken Moffat <[email protected]>: > On Thu, Aug 17, 2017 at 05:29:29AM +0100, Ken Moffat wrote: >> >> I saw more test failures than I was expecting, but some of those >> were probably because I've missed recent fixes for tests. >> > > Heh, trying to fix these up before the next build (I trashed that > build trying to rebuild glibc, but since it wasn't good enough to > build rpcbind it is no loss), I don't think we've got all the errors > covered: > > 1. In acl, my tests crapped out: > > make[1]: Entering directory '/building/acl-2.2.52/test' > *** malformed-restore.test *** > Unescaped left brace in regex is illegal here in regex; marked by > <-- HERE in m/%{ <-- HERE (\w+)}/ at run line 73. > make[1]: *** [Makefile:39: malformed-restore.test] Error 255 > make[1]: Leaving directory '/building/acl-2.2.52/test' > make: *** [Makefile:115: tests] Error 2 > > I _do_ have the sed for the sbits-restore, cp and misc testfiles. > > And in my script I note that some acl tests require mounting with > acl options - is that wrong ? > > A quick test on the perl-5.24 system I'm typing this from gave a > deprecated message (run, line 73) for malformed-restore.test, also for > utf8-filenames.test, sbits-restore.test, setfacl-x.test, and ended > up with 8 tests failed. > > If I add - > > sed -i 's/{(/\\{(/' test/run > > that gets rid of the deprecated messages so I assume it will allow > the tests to complete with perl-5.26, even if some fail. > > 2. In attr, I note that the book forces -j1 for the tests, and my > failures were when running multiple jobs - did something change in > 'make' ? I've noticed one or two BLFS packages recently where > 'make check' automatically picked up $MAKEFLAGS and apparently I've > wrongly blamed the package authors. Back in the day, setting > MAKEFLAGS only made a difference to 'make check' if you specifically > passed the flags to it (i.e. 'make check' defaulted to 1 job). > > 3. I thought I'd botched bison and tried to run the tests before > flex, but I was mistaken and I did run the tests *after* flex. > But they failed - > > make[3]: 'tests/atlocal' is up to date. > make[3]: Entering directory '/building/bison-3.0.4' > YACC examples/calc++/calc++-parser.stamp > make[3]: Leaving directory '/building/bison-3.0.4' > make[3]: Entering directory '/building/bison-3.0.4' > CXX examples/calc++/examples_calc___calc__-calc++-scanner.o > g++: error: ./examples/calc++/calc++-scanner.cc: No such file or > directory > g++: fatal error: no input files > compilation terminated. > make[3]: *** [Makefile:4081: > examples/calc++/examples_calc___calc__-calc++-scanner.o] Error 1 > > Again (point 2 above) I see that parallel jobs were running, I'll > try forcing -j1 here. > -- > I live in a city. I know sparrows from starlings. After that > everything is a duck as far as I'm concerned. -- Monstrous Regiment > -- > http://lists.linuxfromscratch.org/listinfo/lfs-dev > FAQ: http://www.linuxfromscratch.org/faq/ > Unsubscribe: See the above information page -- http://lists.linuxfromscratch.org/listinfo/lfs-dev FAQ: http://www.linuxfromscratch.org/faq/ Unsubscribe: See the above information page
