Micah Cowan <[EMAIL PROTECTED]> writes: > Simon Josefsson wrote: >>> Sadly, old versions of gzip send the help/usage report to stderr, not >>> stdout, so ">/dev/null" alone does not discard it. (And grep wouldn't >>> see an "rsyncable", even if it were there.) Folding stderr into stdout >>> looks harmless to me (not quite a proof of correctness, I realize), and >>> seems to solve the spewage problem: >>> >>> gzip_rsyncable := \ >>> $(shell gzip --help 2>&1 |grep rsyncable >/dev/null && echo >>> --rsyncable) >>> >>> >>> For the record, that's on Tru64 V5.1B-4 (Patch Kit 6), but I'd expect >>> that anyone with a sufficiently obsolete gzip would see it, too. >> >> Fixed in gnulib's maint.mk, thanks! > > Thanks. Did you opt for the 2>/dev/null version because it is known that > no version of gzip old enough to emit --help's output to stderr supports > rsyncable? Otherwise, I'd agree with Steven that the 2>&1 version would > be a bit better.
I synchronized the file with coreutils, which is the "upstream" as far as maintainer-makefile is concerned. /Simon