Subject: | |
From: | |
Reply To: | |
Date: | Sat, 5 Apr 1997 19:33:24 +0200 |
Content-Type: | text/plain |
Parts/Attachments: |
|
|
At 16:43 04.04.1997 -0800, you wrote:
>One of my users playing around with Samba/iX discovered he has some MPE files
>which when accessed from Win95 can cause certain Win95 apps to die or
>experience read errors. But other apps work fine.
>
>He has big files that produce this problem way down in the middle of the file,
>and one small 23-line job stream file that produces the problem immediately.
I haven't checked your sample file but happen to have suffered a similarly
sounding problem when "playing" with a copy of HPSWINFO yesterday evening.
When loading this non-bytestream file into Windows Write (oops, WordPad 95)
I noticed that it had a large amount of garbage characters after the last
line of legible text (all garbage characters were the same, maybe just the
graphical representation of NUL).
I suspect this problem to be rooted (is this proper English??) in the area of
EOF versus "file size perceived when reading" and bytestream-emulation of the
file system (type-manager). Bytestream-emulation can make a record-oriented
text file appear as a sequence of LF-terminated lines to the reader. The LF's
are "faked" and trailing blanks are trimmed...
Compare the values given by ":LISTF,2" and "$ls -l" and "$wc" to get an idea.
If a program first retrieves the "file size" (with stat or fstat??) and then
goes ahead to (raw) read that number of bytes it might get a pretty different
result than a program that reads data until EOF is reported...
I believe I also had some problems to use "diff" on record-oriented files like
HPSWINFO.PUB.SYS in the past, which might be pretty related to this problem...
I'm just scratching the surface so far. Need invest in detail. Help welcome!
Regards, Lars.
PS: Your subject made me fear that Samba had crashed the 3000 (like ping) :-O
|
|
|