#113 parallel builds broken

[Tiziano Müller]

It seems that the build-system is broken in respect to parallel builds ("-jN" parameter for make).
There are random build failures like the following:

*********************************************
[...]
i686-pc-linux-gnu-g++ -pthread -fexceptions -fident -g -fno-inline
-fuse-cxa-atexit -O2 -march=pentium4m -pipe -D_FILE_OFFSET_BITS=64
-D_LARGEFILE_SOURCE -D_LARGEFILE64_SOURCE -D_REENTRANT -D_STLP_DEBUG
-D_STLP_DEBUG_UNINITIALIZED -I../../../stlport -c -o
obj/gcc/so_stlg/test_main.o ../../../test/unit/cppunit/test_main.cpp
Assembler messages:
FATAL: can't create obj/gcc/so_stlg/test_main.o: No such file or directory
make: *** [obj/gcc/so_stlg/test_main.o] Error 2
make: *** Waiting for unfinished jobs....
make: Leaving directory
`/var/tmp/portage/dev-libs/STLport-5.1.0/work/STLport-5.1.0/build/test/unit'
*********************************************

This happens when building the tests as well as when building the library itself.

[Francois Dumont]

Logged In: YES
user_id=1096600
Originator: NO

Just for your info, I have tested the -j make option under cygwin and it works like a charm. I would be surprised that GNU make files need to respect any special rule in order to make -j option work. Are you sure problem is not in your make command, what version are you using ?

[Tiziano Müller]

Logged In: YES
user_id=935305
Originator: YES

Well, the problem of parallel builds is that it isn't necessarily reproducible. On my machine everything works ok, on a Co-Dev's machine (Intel Core 2 Duo, GNU/make-3.8.1) it always fails on different locations until he disables parallel builds.

[Petr Ovtchenkov]

Logged In: YES
user_id=615813
Originator: NO

Don't see problem as on slow uni-processor matchines, as on very fast 4-proc SMP hosts; gmake was as 3.80 as 3.81. Resume: check you OE, hardware, shell, gcc or make. Note, that parallel build may depends upon you shell etc.