Re: mkdir doco

[Alexandre Duret-Lutz]

>>> "Jim" == Jim Meyering <address@hidden> writes:

[...]

 Jim> One minor correction: I think that bug in GNU mkdir was fixed
 Jim> in fileutils-4.1-b2, which was followed by fileutils-4.0c.

Thanks!  Maybe it's enough if I write just `GNU Coreutils', since
all versions of GNU Coreutils are OK.  Or shall I say
`GNU Coreutils (since Fileutils version 4.0c)'?

 Jim> FWIW, the bug was exposed when I saw parallel builds failing
 Jim> because `mkdir -p lock-directory' would sometimes fail.
 Jim> Here's the relevant ChangeLog entry:

 Jim> 1999-01-01  Jim Meyering  <address@hidden>

 Jim> * lib/makepath.c (make_dir): New function, factored out of make_path.
 Jim> (make_path): Use make_dir rather than open-coding it twice.
 Jim> This effectively reverses the order of the latter pair of stat/mkdir
 Jim> calls and fixes a race condition bug whereby one of two concurrent
 Jim> `mkdir -p' processes could fail with EEXIST.

I stopped when I found this one:

Tue Jul  1 06:42:17 1997  Jim Meyering  <address@hidden>

        * lib/makepath.c (make_path): Reorder stat-then-mkdir-if-missing
        calls so that mkdir is called first.  Before make_path would first
        `stat' a directory, then call mkdir if it didn't exist.  But if
        some other process created the directory between the stat & mkdir,
        the mkdir would fail with EEXIST.  Diagnosis and suggestion from
        Greg McGary.

-- 
Alexandre Duret-Lutz