On 07/13/2014 03:44 PM, Pádraig Brady wrote: > On 07/13/2014 01:59 PM, Bernhard Voelker wrote: >> On 07/12/2014 06:56 PM, Pádraig Brady wrote: >>> Here is the complete rolled up patch, which I'm about ready to push. >>> http://www.pixelbeat.org/cu/single-binary_v10.patch >> >> Thanks! >> >> The man pages issues are almost gone now: the only difference >> (compared to an installation from master) still is in man1/test.1: >> >> - Report test translation bugs to <http://translationproject.org/team/> >> + Report [ translation bugs to <http://translationproject.org/team/> > > I left that, as it shouldn't cause user confusion > and would be a bit awkward to fix. > >> Another minor nit in the commit message: >> >>> * .gitignore: Added new generated files. >> >> s/: Added /: Add / (4 matches). >> >> Otherwise +1 ;-) > > Adjusted those and pushed. > > thanks for all the reviews!
BTW I see some tests being skipped in single binary mode, due to ulimit issues. It seems like something like the following will be needed, along with parameterized values for the base limit to use in all the ulimit using tests. I don't see that as a requirement for the upcoming release though but we might add it if we need to respin the snapshot for some reason. thanks, Pádraig. diff --git a/init.cfg b/init.cfg index 725ee12..fb6890c 100644 --- a/init.cfg +++ b/init.cfg @@ -157,6 +157,12 @@ require_ulimit_v_() local vm case $(printenv LD_PRELOAD) in */valgrind/*) vm=22000;; *) vm=10000;; esac + # The single binary requires much more virtual memory to map + # in all the shared libraries etc. + if test -s "$abs_top_srcdir/src/coreutils.h"; then + vm=$(expr $vm + 15000) + fi + ( ulimit -v $vm; date ) > /dev/null 2>&1 || ulimit_works=no ( ulimit -v 20; date ) > /dev/null 2>&1 && ulimit_works=no