Hi Adam -
There is an old issue having to do with the windows
version where to prevent an endless loop of trying to process a file that errors
out and cannot produce a corresponding css with matching timestamp. If an
error is encountered, the watch process stops watching / temporarily
ignores a file with a timestamp still matching the timestamp when an error
was detected (or at least I think that was the design intention.) In
practice, it seems to be killing the watch after one iteration whether an error
is reported or not - perhaps the success message is being interpreted as an
error. I'm not on the latest version so I don't know it this is still
the case, I know that the version I am running aborts the watch when a file
change is detected and has to be manually restarted after each
revision.
I'll update later this evening and see if the issue
persists. I know it's been on Chris' list for several months and it's not
always convenient for him to reproduce and troubleshoot windows-only
issues.
Duke