Casper.Dik at Sun.COM wrote:
>
> >Fixed - see below...
> >
> >> - this is caused by calling:
> >>
> >> cmd=$(basename "$0")
> >>
> >> - and I noticed that it didn't happen when I use:
> >>
> >> cmd=${0##*/}
> >
> >Mhhh, this works because there is no PATH lookup required to find the
> >matching command...
>
> That's not the entire story.
>
> basename "$0" - no stat'ing .paths
> cmd=$(basename "$0") - stating .paths
Uhm... right now I have no clue... I have to crawl around in the source
a bit (it seems that changing $(...) to ${...;} causes lookups for
".paths", too - which rules out that the subshell code is causing this)
...
> >The idea was to rename the PATH variable (e.g. via $ typeset -m ... #)
> >to something like "saved_path" and later do a $ PATH="${saved_path}"
> >"${cmd}" "$@" # which would have a similar effect...
> >... but your solution is shorter (and works since "sleep" is not bound
> >to a specific PATH element).
>
> And the other commands alias'ed to "alias"?
Seems to be fixed for them, too:
-- snip --
$ (for i in sleep kill test alias printf command alias fg bg jobs ; do
print "## $i" ; rm -f /tmp/yyy ; truss -o /tmp/yyy $ROOT/usr/bin/jobs ;
fgrep ".path" </tmp/yyy ; done)
## sleep
## kill
## test
## alias
## printf
## command
## alias
## fg
## bg
## jobs
-- snip --
What else needs to be fixed (excluding [1]) ?
[1]=CR #6807629 ("sleep: time argument should ignore LC_* variables")
may require ask the Austin Group how to interpret the POSIX/SUS standard
in this case. IMO "sleep"'s argument should use LANG/LC_NUMERIC/LC_ALL
(in that order) for parsing it's arguments like |libc::strtod()| and
|libc::sscanf()| do (otherwise there is the question how a shell with
floating-point support for ((<arithmetric-expression)) should look at
LANG/LC_NUMERIC/LC_ALL, too (which ksh93 does)).
----
Bye,
Roland
--
__ . . __
(o.\ \/ /.o) roland.mainz at nrubsig.org
\__\/\/__/ MPEG specialist, C&&JAVA&&Sun&&Unix programmer
/O /==\ O\ TEL +49 641 3992797
(;O/ \/ \O;)
