I have have reported this before. I figured it would take all of ten minutes to fix, but it is still there. I don't think I jumped up and down enough about how important a bug it is to fix. Let me explain the problem again.
There are several ways the problem manifests:
1. I load htmlvalidator when I am in directory A.
2. I then leave it running.
3. then I move to another directory B and invoke a command line validation of a file C which lives in dir B.
It won't be able to find it. It looks in the original dir A, not the current dir B.
Another is it asks me if I want to reload because I edited the file with some tool outside of Htmlvalidator, I say yes, but it fails presumably because the current directory has changed since HTMLValidator loaded.
Another case. A script that loads and edits a file will work if HTML validator is not loaded, but will fail it if is already loaded. My workaround is to keep killing HTML validator, knowing perfectly well I will load it a again a few seconds later. By I keep forgetting and my scripts fail.
The fix is simple. Every editor does this. It must track files internally with the absolute filename, not the relative one of any files it has loaded.
When you run an HTMLValidator command line, it must use the CWD, not the CWD when the bulk of it was loaded earlier.
I think simply storing all filenames internally in absolute form will clear up most of the trouble.



