Re: [patch@35055] vms.c rmsexpand refactor

Craig A. Berry wrote:
> 
> On Dec 7, 2008, at 11:35 PM, John E. Malmberg wrote:
> 
>> Craig A. Berry wrote:
> 
>> It may be easier to apply the patches and then compare the results 
>> against the unpatched vms.c manually.
> 
> Not really, and that's not something a committer should be asked to do 
> anyway.  I've just spent over an hour trying to get this series of six 
> "refactor" patches to apply, but no luck.  There were six separate 
> patches for pathify_dirspec, unixify, unixpath, vmsify, fileify, and 
> rmsexpand.  Here and there a hunk succeeded but most did not.  The hunks 
> that did apply had hundreds of lines of offset, indicating something's 
> missing.

I do not know what happened there, I set up multiple reference 
directories so that each patch was incremental.

The order was important because of name changes to the internal routines.

> I fiddled with the order in which I applied them, used -l, and spent 
> some time trying to figure out what was what before giving up.  Any 
> manual attempt to sort out what was going on was pretty much sunk by 
> having hundreds of lines consisting only of non-significant whitespace 
> changes -- consider using diff -b.

I made the whitespace changes in order to make the code easier to follow 
while I was refactoring it.

How about I submit a patch that does the whitespace changes only as a 
first step with no other bug fixes?

And to make things consistent, what tab character usage policy do you 
want?  No tabs, tabs on 8 character boundaries?

I prefer the tabs on 8 character boundaries.

> I look forward to a revised patch that applies cleanly and does not 
> introduce new test failures.

Since submitting them, I have found more bugs in VMS.c, some of which 
are in the patched routines.

Before resubmitting I have one more test to get running in UNIX mode to 
get working.

And the POSIX exit code setting still has some bugs in it.

There is also a bug in the pipe exit handler, it is access violating if 
there is a pipe open to a child.  I am only seeing this while debugging, 
and using the EXIT command in the debugger.

-John
[email protected]
-- 
The [email protected] and my encompasserve.org e-mail addresses are out of 
service for the moment.