"extract" script for extraction. Many cases where a custom EXTRACT_CMD
simply copied the distfile into the work directory are no longer
needed. The extract script also hides differences between pax and
tar behind a common command-line interface, so we no longer need code
that's conditional on whether EXTRACT_USING is tar or pax.
needed for troubleshooting). And use BINOWN and BINGRP because user "bin"
and group "bin" don't exist on some systems.
Use BSD_INSTALL_PROGRAM, BSD_INSTALL_DATA and BSD_INSTALL_SCRIPT
instead of calling /usr/bin/install directly.
Okayed by maintainer.
Bumped PKGREVISION, since ownership of files may be different on
some systems.
It is more consistent with the tex.buildlink3.mk name. Also, if a package
really needs latex, it just has to set TEX_ACCEPTED to latex distributions
altough today, all TEX_ACCEPTED possibilities are latex distributions
fix the signal handling to use a volatile variable. Compilers can
optimise the access otherwise. Bump revision, since this is a bugfix.
Reviewed by jlam and the maintainer.
directories to process. The removal of pre-configure fixes the config.*
handling at the same time. Also specify the config.status overrides.
Patch configure to not try to detect -lrfftw, just link the static
version explicitly. Do the same for -lfftw. This ensures that we
actually do get the static versions without having to worry about
libtool being too clever. Retire manual config.sub hacks, this should be
handled by config.guess already.
<jlam> just go ahead and commit that benchfft patch and let's
<jlam> never speak of it again :)
While there, use fsync(2) on all hosts. Without this, the fstime benchmark
will be disproportionately biased toward OS's and hardware that buffer
larger amounts of data in memory before [background] syncing.
Classic Unix fork() bomber. Includes CPU hanger, memory allocator,
memory toucher and zombie dance team. You can test how will your
computer behave under heavy CPU, process, memory load.
Forkbomb is also useful as realloc() benchmark.
