PR/58577 CVS commit: src/usr.bin/xinstall

["Robert Elz" <kre%netbsd.org@localhost>]

The following reply was made to PR bin/58577; it has been noted by GNATS.

From: "Robert Elz" <kre%netbsd.org@localhost>
To: gnats-bugs%gnats.NetBSD.org@localhost
Cc: 
Subject: PR/58577 CVS commit: src/usr.bin/xinstall
Date: Fri, 10 Jul 2026 07:55:51 +0000

 Module Name:	src
 Committed By:	kre
 Date:		Fri Jul 10 07:55:50 UTC 2026
 
 Modified Files:
 	src/usr.bin/xinstall: xinstall.c
 
 Log Message:
 PR bin/58577  - install(1) -d issues
 
 Fix issues where "install -d" (with no directory) simply
 exit(0)s.   That one is kind of marginal, installing nothing
 when nothing is needed could be treated as OK, but the man
 page does indicate in the SYNOPSIS that with -d, at least
 one directory is needed (it says nothing at all about that
 in the text).
 
 Second, after creating a directory, if a later operation
 (chown, chmod) fails, that is not success, a warning was
 issued (good), a bad metalog was being created (bad).
 That is clearly a bug (though probably doesn't happen
 very often).
 
 And third, when install -d fails, and issues an error,
 (warn()) the exit status from install(1) should not be zero.
 This only applies after the (rare, I'd assume) errors
 from the 2nd fix (which were not previously regarded as
 errors).  The more common failure cases were before, and
 still are, handled by an immediate err() exit.
 
 Patch from the PR (from RVP) installed, plus a couple of
 changes from me.
 
 The method in the patch of just "or"ing the status from
 the install_dir() would work if install_dir() had been made
 to return 0 (success) and 1 (failure), but it wasn't, it
 was made to return EXIT_SUCCESS and EXIT_FAILURE instead.
 It is wrong to assume those are necessarily 0 and 1.
 They are for us (and other unix like systems), but not
 guaranteed.   So I modified that code to eliminate that
 coding technique (I think when creating directories, we
 can afford one more "if" in the code!)
 
 I also added sets of {} around a few large blocks of a single
 statement that just looked weird (not incorrect) as they were.
 This thing still needs some KNF surgery however (including, but
 not limited to, whitespace issues).
 
 XXX pullup -11  (probably -10 as well)
 
 
 To generate a diff of this commit:
 cvs rdiff -u -r1.130 -r1.131 src/usr.bin/xinstall/xinstall.c
 
 Please note that diffs are not public domain; they are subject to the
 copyright notices on the relevant files.