Return-Path: 
 <dev-return-8770-apmail-apr-dev-archive=apr.apache.org@apr.apache.org>
Delivered-To: apmail-apr-dev-archive@apr.apache.org
Received: (qmail 99842 invoked by uid 500); 18 Dec 2002 14:14:36 -0000
Mailing-List: contact dev-help@apr.apache.org; run by ezmlm
Precedence: bulk
List-Post: <mailto:dev@apr.apache.org>
List-Help: <mailto:dev-help@apr.apache.org>
List-Unsubscribe: <mailto:dev-unsubscribe@apr.apache.org>
List-Subscribe: <mailto:dev-subscribe@apr.apache.org>
Delivered-To: mailing list dev@apr.apache.org
Received: (qmail 99828 invoked from network); 18 Dec 2002 14:14:35 -0000
To: "William A. Rowe, Jr." <wrowe@apache.org>
Cc: Branko =?iso-8859-1?q?=C4=8Cibej?= <brane@xbc.nu>,
  dev@subversion.tigris.org,
	  dev@apr.apache.org
Subject: Re: [PATCH] apr_dir_read doesn't return requested information
References: <3DFCC43F.F6B93489@sebastian-bergmann.de>
	<87fzsz6o1o.fsf@codematters.co.uk>
	<3DFD61A3.D6C1DFF@sebastian-bergmann.de>
	<87znr60wpx.fsf@codematters.co.uk>
	<3DFE128D.14A41D65@sebastian-bergmann.de>
	<87pts122qe.fsf@codematters.co.uk>
	<3DFE4431.9F5DA96E@sebastian-bergmann.de>
	<8765tt1qzo.fsf_-_@codematters.co.uk>
	<87k7i9zcbc.fsf_-_@codematters.co.uk>
	<5.1.0.14.2.20021217215518.03670928@pop3.rowe-clan.net>
	<5.1.0.14.2.20021217225946.02a1ae90@pop3.rowe-clan.net>
From: Philip Martin <philip@codematters.co.uk>
Date: 18 Dec 2002 14:14:29 +0000
In-Reply-To: <5.1.0.14.2.20021217225946.02a1ae90@pop3.rowe-clan.net>
Message-ID: <87bs3jct2y.fsf@codematters.co.uk>
Lines: 79
User-Agent: Gnus/5.0808 (Gnus v5.8.8) XEmacs/21.4 (Common Lisp)
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Sender: Philip Martin <pm@localhost>
X-Spam-Rating: daedalus.apache.org 1.6.2 0/1000/N

"William A. Rowe, Jr." <wrowe@apache.org> writes:

> What I would like to know (if you can track it...) 
> 
> Is it possible to dump the finfo structure within gdb at the point this
> request fails?  I'd pay especially close attention to the .valid bits, since
> those are the identifiers that will help us determine if stat() was also
> called later.

This is for dir.c version 1.71 with the patch reverted.  The
Subversion code is svn_io_get_dirents in subversion/libsvn_subr/io.c,
it passes APR_FINFO_TYPE | APR_FINFO_NAME to apr_dir_read.  The first
two calls to apr_dir_read return "." and ".." and the Subversion code
skips them, the following gdb information is for the third call

(gdb) s
apr_dir_read (finfo=0xbffff660, wanted=33587200, thedir=0x809a878) at dir.c:174
174         apr_status_t ret = 0;
(gdb) n
179         ret = readdir_r(thedir->dirstruct, thedir->entry, &retent);
(gdb) 
184         if(!ret && thedir->entry != retent)
(gdb) p ret
$1 = 0
(gdb) p thedir->entry[0]
$2 = {d_ino = 186434, d_off = 13512064, d_reclen = 16, d_type = 0 '\0', 
  d_name = "..", '\0' <repeats 253 times>}
(gdb) n
194         if (ret == EINVAL) {
(gdb) 
214         finfo->fname = NULL;
(gdb) 
216         if (ret) {
(gdb) 
222         wanted &= ~APR_FINFO_INODE;
(gdb) p/x wanted
$3 = 0x2008000
(gdb) n
225         wanted &= ~APR_FINFO_TYPE;
(gdb) 
228         wanted &= ~APR_FINFO_NAME;
(gdb) 
230         if (wanted)
(gdb) p wanted
$4 = 0
(gdb) n
244         if (wanted && (ret == APR_SUCCESS || ret == APR_INCOMPLETE)) {
(gdb) 
251             finfo->pool = thedir->pool;
(gdb) 
252             finfo->valid = 0;
(gdb) 
254             finfo->filetype = filetype_from_dirent_type(thedir->entry->DIRENT_TYPE);
(gdb) 
255             finfo->valid |= APR_FINFO_TYPE;
(gdb) p finfo->filetype
$5 = APR_UNKFILE
(gdb) n
258             finfo->inode = thedir->entry->DIRENT_INODE;
(gdb) 
259             finfo->valid |= APR_FINFO_INODE;
(gdb) 
263         finfo->name = apr_pstrdup(thedir->pool, thedir->entry->d_name);
(gdb) 
264         finfo->valid |= APR_FINFO_NAME;
(gdb) 
266         if (wanted)
(gdb) 
269         return APR_SUCCESS;
(gdb) 

Subversion explicitly requests APR_FINFO_TYPE but then the
apr_dir_read code clears that bit and so doesn't call apr_lstat.
APR_UNKFILE appears to be correct for d_type of 0, but is not a very
useful thing for APR to return in response to APR_FINFO_TYPE.  It's a
change in APR's behaviour and it breaks Subversion.

-- 
Philip Martin