Subject: Re: changed files: 'src/sys/miscfs/procfs procfs_vnops.c'
To: None <,>
From: Wolfgang Solfrank <>
List: source-changes
Date: 02/15/1994 16:49:27
> > This change is wrong.  First, it's gratuitous.  Second, it ignores the
> > case of other executable formats.
> i agree with charles that changing 'file' to 'a.out' in this case
> is a bad idea; it ignores other executable formats to save 5
> characters of typing to gdb.

I didn't even think about saving some typing in gdb, but that the name
'file' didn't give any information what this is about. It's simply ugly.

And the case of other executable formats is obvisouly a non-issue, since
every U*x derived system I have access to (SunOS, Solaris, NEXTSTEP, Xenix,
SCO Unix) with so different executable formats as a.out, elf, mach-o, xenix
modified a.out and coff all generate their executables with the name a.out,
if you call the compiler without explicitly giving a name.

BTW, IMHO "mem" has to change, too. Currently it gives file sizes
that are rather different from the possible seek offsets. Seek offsets
are interpreted as virtual addresses in the process' address space, but
the size of the file is given as the total of text, (current) data and
stack size.

Either the size given should be VM_MAXUSER_ADDRESS (problematic, since
this would be interpreted by most programs as a negative number, which
BTW IMHO is the longest standing bug in U*X), or the seek offsets should
be interpreted similar to the size computation. In the latter case the
file contents should be interpreted similar to the format of core dumps,
and the file should be renamed "core".

> also, it has a potential for weirdness when combined with 'find's of
> old a.out's...

This may be an issue, but then may be not. IMHO, you should have to stop
such finds from traversing filesystems it doesn't know about...
ws@TooLs.DE     (Wolfgang Solfrank, TooLs GmbH) +49-228-985800