Where I’m @

20040329

TortoiseCVS/CVSNT Notes, Three
Finally managed to get the postcommit script working, and here’s what we learned, not necessarily in the order in which we learned it:

• In the CVSNT postcommit file, the location for the postcommit script must be a full Windows path if you are going to use the Windows Scripting Host (WSH). The CVSNT Wiki sample postcommit script uses the environment variable $CVSROOT to begin the script location path, but we found that that confuses WSH.
• In a multiple file check-in, all files are committed before the postcommit script runs, so following the sample script to the letter means that only the last committed file’s directory will get passed as a parameter to the postcommit script, and that is the only directory that will get updated. If you only check in files from one directory at once, that’s fine, but we sometimes want to be able to check in multiple files from different directories and have the whole project update. So we made a copy of the script that strips all but the root directory from the check-in path, which means that the entire project will get updated at once^*. We also use a version of the same script that only updates the directory that the committed file is from, again based on the postcommit config file.
• We discovered that there is a way to create a project directory that has no CVSNT admin files or directories in it: Use the Option tab in the Checkout dialogue, and select “Export” instead of “Checkout” under “Purpose of checkout”. (What is this, customs?)
• However, it’s better to use checkout rather than export with the postcommit script. The main functional part of the script is to call a CVSNT command from the command line, and export has no options to remove or add directories. In other words, using checkout makes the postcommit script essentially a mirror script, which export is a snapshot of sorts. It would seem possible that pushing files to production might be automated using export, but that might cause problems with datestamps.

Our CVS repository path is drive:\cvsrepo\[project] and our development path pattern is drive:\[dev directory]\[project]. So when a file is checked in, the parameter passed is \cvsrepo\[project]\[dir]\[file]. We use the essentials of the CVSNT wiki’s sample script to strip off \cvsrepo and \[dir]\[file] and to add drive:\[dev directory] so what gets passed to a commandline as the checkout directory to update is the development path pattern — that is, the root directory of the dev project.

20040326

TortoiseCVS/CVSNT Notes, Two
The best way to get answers for CVSNT is to search the mailing list using our friend Google, or to ask the list yourself. There are three or four people who seem to have most of the best answers, as seems often to be the case on lists of this sort, but stick around — you may be able to help out eventually.

TortoiseCVS/CVSNT Notes, One
To make changes to CVS administration files, you have to check out/edit/commit the files the same way you check anything else. Learning this (from an email in the CVSNT newsgroup/mailing list archives) helped me a whole bunch. We’re working on getting something set up for a postcommit script, and it’s driving us crazy. However, we have been able to get one script working, so it’s getting brighter.

20040323

Caution: Learning Curves Ahead
We are learning a number of new things in our new situation, the current one being CVSNT and the TortoiseCVS client. Here are the problems we are having so far:

• How does a new file get added into CVS by a developer who doesn’t explicitly have write access to the development server? It’s clear how a file gets registered in CVS, but how does the file get moved across the filesystem from the developer’s local machine to the dev server’s CVS repository?
• Somewhat related, where are the files that can be pushed to production? We’re not working with compiled software, just old-fashioned ASP, and we need to make sure we are pushing the latest files to the staging webserver and then the production webserver.
• Is there a way through Tortoise to see who’s editing which files?
• One of our developers is wary of running IIS on her box. Is there any way she can work over the network on a special repository? (We are pretty sure the answer is no, but we aren’t 100% there.)

We have no fear of exposing our ignorance in public, as you can see, and we are doing research to find the answers to these questions. Thus far in our career, we’ve only run CVS on our box at home and have not had to worry about multi-user issues. As a whole, these issues can generally be filed under the broader question, “What is the relationship between the filesystem and CVS file tracking?” Once files are committed into CVS, can we delete the original files, as we have done with Microsoft’s Visual Source Safe? How are files centralized? How do projects such as TortoiseCVS itself manage developers scattered geographically? Can they all add files?