RE: PR 41130

["Larson, Timothy E." <TELarson%west.com@localhost>]

> > I can't yet report that bootstrapping works on A/UX, but I am
> > certain that the shell-related hacks to the bootstrap script are
> > definitely needed because of the shortcomings of A/UX's shell.
> > There isn't any impact to any other OS, so these should be quite
> > safe to commit.
> 
> "check_prog shprog ksh" could be moved to AUX) case part
> (next "check_prog" with same 1st arg will be skipped).

OK, I didn't know check_prog worked that way, and assumed the shprog selection 
would be changed if I'd set it earlier.

> About re-exec part, I can't understand meaning of the comment:
>   "don't do this earlier, because A/UX mkdir will choke on the dir
> creation"

A/UX's mkdir doesn't honor -p, so initially bootstrap dies twice when the work 
subdirs are created IIRC.  (Third try's the charm.)  Re-execing too early will 
erroneously cause the AuxKsh marker file to be present when it's not actually 
running as ksh yet.  

If all instances of "mkdir -p" could be put in a wrapper function to simulate 
the intended behavior, this could probably be moved to much earlier in the 
script.  At one point I'd tried that, but couldn't test my solution 
extensively, so fell back to the workaround of running bootstrap three times.

Tim
-- 
Tim Larson
App Admin, Intercall Production Services