Han Xin <hanxin.hx@xxxxxxxxxxxxx> writes:
> Maybe I should use "lib-bash.sh" instead of "test-lib.sh" just like t9902
> and t9903?
> The different meanings of "-p" in bash and dash really make this tricky.
I do not think lib-bash.sh is appropriate for this.
t9902/9903 are about command line prompt and completion support
_FOR_ bash users. By including lib-bash.sh, even if your "usual"
shell is not bash, you can run these two scripts under bash, as long
as you have it installed.
For the purpose of testing these bash specific features, that
framework makes quite a lot of sense. Those who are happy to have
dash on their system without having to install bash would have no
reason to see these two tests to pass, as they do not care about
bash at all.
What the test under discussion is doing is quite different. Instead
of forcing to re-spawn bash when the user's shell is not bash, you'd
want to adjust how you invoke "ulimit" if it is not bash, something
like
run_with_limited_processes () {
(ulimit ${ulimit_max_process-"-p"} 512 && "$@")
}
test_lazy_prereq ULIMIT_PROCESSES '
# bash and ksh use "ulimit -u", dash uses "ulimit -p"
if test -n "$BASH_VERSION"
then
ulimit_max_process="-u"
elif test -n "$KSH_VERSION"
then
ulimit_max_process="-u"
fi
run_with_limited_processes true
'
perhaps?
