276587 - [BeOS] bfile.c still uses the broken access function for _MD_access

Status: RESOLVED FIXED

(NSPR :: NSPR, defect)

Description

[tqh]

See https://bugzilla.mozilla.org/show_bug.cgi?id=169506#c0 for info.

Comment 1

[tqh]

Attachment: Using stat() instead

Here is a rewrite sort of like in bug 169506.

Comment 2

[Wan-Teh Chang]

Comment on attachment 169976 [details] [diff] [review]
Using stat() instead

>+			//If stat doesn't return error we should never set rv to -1, so st_mode & checkFlags must be > 0.

Please use C-style comment delimiters /* ... */
and wrap this long line.

Is it possible for a file to have a buf.st_mode of 0?
If so, we'll need to handle PR_ACCESS_EXISTS as a special
case without using checkFlags.	I think it's safer to do
that anyway.

>+	rv = stat(name, &buf);
>+	if(rv==0 && ( !(buf.st_mode & checkFlags)))
>+		rv = -1;
> 
> 	if (rv < 0) {
> 		err = _MD_ERRNO();
> 		_PR_MD_MAP_ACCESS_ERROR(err);
> 	}

We should use _PR_MD_MAP_STAT_ERROR instead of
_PR_MD_MAP_ACCESS_ERROR.

If we set rv to -1 artificially, we can't use errno
because the stat() call succeeded.  So we will need
to also set the error code artificially.  This means
the code should read:

	rv = stat(name, &buf);
	if (rv==0 && ( !(buf.st_mode & checkFlags))) {
	    PR_SetError(PR_NO_ACCESS_RIGHTS_ERROR, 0);
	    return -1;
	}

Comment 3

[Wan-Teh Chang]

Attachment: Using stat() instead, v2

This is what I meant.  Please review and test this patch.

Comment 4

[tqh]

Attachment: Same patch but without the \r newline that is bad for BeOS

Except that it had bad EOL characters, it looks ok. Uploading the same patch
but without \r.

Comment 5

[tqh]

Comment on attachment 170187 [details] [diff] [review]
Using stat() instead, v2

Removing obsolete review.

Comment 6

[Wan-Teh Chang]

I checked in the patch on the NSPR tip (NSPR 4.6)
and NSPRPUB_PRE_4_2_CLIENT_BRANCH (Mozilla 1.8 Beta).