I put some timers around the methods in DepTable which re-compute the CRC of each file, and I found it was taking a significant amount of time on my thesis... 30-60 seconds. So I started wondering about optimising it somehow.
At first I wondered why it calculates CRCs instead of checking file modification date/time, but then I realised that the files are usually re-written every time you run latex or bibtex or whatever, so we need to detect if the file contents have changed regardless of the modification date/time. However, not all files are changed - eg input files like .tex and .bib, but they still need to be in the dep table for subsequent runs. For such files, the modification date/time would remain the same (untouched) within one set of runs. So, I would expect that if you were to check the date/time and it had not changed, then you would not need to re-calc the CRC. Of course, if the date/time has changed since the CRC was last calculated, you still have to re-calc the CRC. Any objections? I intend to implement this in my tree - you can have patches after you end your feature freeze. Ben Stanley
