On Sunday, January 13, 2013 04:54:59 PM Chet Ramey wrote: > On 1/9/13 2:00 PM, Dan Douglas wrote: > > When expanding simple commands, steps 3 and 4 are reversed unconditionally > > for > > all command types and number of words expanded, even in POSIX mode. > > http://pubs.opengroup.org/onlinepubs/9699919799/utilities/V3_chap02.html#tag_18_09_01 > > True. Bash has always behaved like this. It's not clear what practical > consequences it has, other than being technically non-standard, since > redirections do not have access to variables set in the temporary > environment.
Yeah... I think I noticed this in trying to figure out what goes on with the environment of a redirection. There's that "command" thing with IFS a few mails back. And then there are some other leaky bits. I ran a bunch of experiments months ago while trying to document the expansion process in more detail. Just one example still with the most recent patch: $ bash -c 'x=1; x=2 true <&0$(eval echo \$x $x >&2)' 2 1 I think you solved some problems in the last patch (the IFS in a process substitution in a redirect glitch). And some probably have to do with all this business: http://article.gmane.org/gmane.comp.standards.posix.austin.general/1927 (I had to re-read that discussion several times.) -- Dan Douglas
