Date: Thu, 25 Jan 1996 20:13:29 CST (-0600)
To: djgpp AT delorie DOT com
From: cameronbuschardt <c027319 AT email4 DOT starnetinc DOT com>
Subject: Re: V2 alpha 960124

>Date: Thu, 25 Jan 1996 15:18:52
>To: Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>
>From: cameronbuschardt <c027319 AT mailc DOT starnetinc DOT com>
>Subject: Re: V2 alpha 960124
>
>At 06:16 PM 1/25/96 +0200, you wrote:
>>
>>On Thu, 25 Jan 1996, DJ Delorie wrote:
>>
>>> Hmmm.. I think my subconsious is trying to program also.  I feel that
>>> access() is used enough during startup to try to avoid keeping
>>> findfirst from linking in.  stat() handles root directories already;
>>> can we switch opendir back to the old way (call findfirst) rather than
>>> call access?
>>
>>If you don't disable filename globbing, `findfirst' is called during
>>startup anyway (by `fnmatch'), isn't it? 
>>
>>`opendir' can indeed be fixed by restoring the old code (unless you had 
>>good reasons to change it), but I don't think we can live with `access' 
>>which fails on some root directories.  For instance, Emacs calls `access' 
>>to see whether a directory exists (is readable) before running Dired on 
>>it.
>>
>>The real problem is that `access' calls `_chmod', which isn't guaranteed 
>>to work on root directories because they aren't true files.  Is there 
>>another, cheaper way to detect a root directory of a valid drive?
>>
>>If this problem is limited to CD-ROM drives (I didn't see it with any
>>other drives--Novell, NFS, RAM etc.), then a small function
>>`is_cdrom_drive' from src/libc/compat/mntent/mntent.c might be all we
>>need.  Can other people test `access' with all the network drives you can
>>reach and verify that this problem is limted to CD-ROM? 
>
>There are 3 possibilies to detect if a drive is valid.
>1. Record waht the current drive is, then switch to the drive in question.
>   Next check to see if there were any errors if not then switch back to the
>   original drive.
>
>2.) Use  Int 21h,  Ah=247h (Get current dir on specified drive.
>    where:
>        DS:SI points to 64 byte buffer (for dir)
>        DL=drive number (0=a 1=b..) (Note 2 is not drive C)
>    Return:
>        Carry flag set if error 
>        ah=15 if INVALID drive specification.
>
>3.) Use BIOS.
>There is probably another way but these are the only ones I know of.
>