Re: -t/-f in cvswrappers

[DAve Goodrich]

on 4/13/01 12:26 PM, address@hidden at address@hidden
wrote:

>> on 4/12/01 4:36 PM, Larry Jones at address@hidden wrote:
>>> DAve Goodrich writes:
>>>> 
>>>> Hello all. I'm running into a problem with clients that send files to the
>>>> repository with Mac or Win file endings.
>>> 
>>> What clients are these?  It is the client's responsibility to convert
>>> text files from whatever the local line ending convention is to the
>>> canonical line-feed termination before sending them to the server.  If
>>> that isn't happening, the clients are buggy or misconfigured.
>>> 
> 
>> We have several people using MacCVSClient, which uses Mac line endings by
>> default. The developer knows that people who are creating code that must
>> eventually live on a Unix box(web pages) have a problem with this (me) and
>> he is working to remedy that.
> 
> the line endings should get translated by the client.  If they are not
> being translated, then there is a problem with that client.  Perhaps it
> is misconfigured.
> 
Agreed, the developer know this and is working on it. See above.

>>> You
>>> aren't using some kind of shared file system and trying to commit files
>>> from a different platform than the one you're running on, are you?
> 
>> Nope creating web pages on Mac/Windows and serving from Unix.
> 
> I do not think you understood his question.  Are you using the pserver
> method, the rsh/ssh method or a shared drive to access the cvs
> repository.  If you are using a shared drive, then this is your
> problem.
> 
Sorry, I didn't understand his question. We are using pserver as we are
inside a firewall without outside access. The files themselves are moved via
cvs export to a staging server, then rsync'd via ssh to the production box.

The issue is only with developers committing files via pserver.

>>>> But I get an error that "cvs [server aborted]: -t/-f wrappers not supported
>>>> by this version of CVS"
>>> 
>>> There were bugs in the -t/-f wrapper code that could cause silent loss
> 
>> Thanks, not what I wanted to hear, but oh well.
> 
>> Seems that would be an extremely useful function. I would like to be able to
>> filter many things through a file prior to committing, check for common
>> copyright code at the bottom of a file, check for tags ($RCSfile$) at the
>> beginning of each file, things I must do weekly in code review. Odd it seems
>> to have been forgotten so long.
> 
> All these things can be done using a trigger script placed in the commitinfo
> file.
> 
> Rex.

I'll look into that, any examples? I'm not sure how that would work. Would
this enable a file to be changed? or would it only be able to validate and
then reject/accept a file for commit based on the result of the regular
expression. Ideally to solve this problem, I need to filter a file through a
process between the client and the repository.

Thanks

DAve

--
Dave Goodrich
Director of Interface Development
Reality Based Learning Company
9521 NE Willows Road, Suite 100
Redmond, WA 98052 
Toll Free 1-877-869-6603 ext. 237
Fax (425) 558-5655 
address@hidden 
http://www.rblc.com