Filesystem scan doesn't cache on failed build

[Qix-]

Upon a lengthy filesystem scan followed by a failed build, the filesystem results are not saved thus prompting a long filesystem scan each time.

For instance, I'm running into multiple output conflicts, and each time I need to test (or even just run tup ref) it insists on running the filesystem scan again.

Could the results be added to the tup repository prior to doing anything else so it turns into a one-time thing?

[gittup]

Are you talking about the time it takes to filesystem scan, or the fact that Tupfiles are re-parsed if any of them fail? ie: is "time tup scan" a really long time? You could pass in '--no-scan' if you are certain nothing has changed in that case. Saving the completed Tupfile parsings when one fails is another issue, though.

[Qix-]

If I recall what was happening with this problem, here are the events that occur:

1. A really long scan time (like 2 minutes; lots of files)
2. Scan completes fine
3. Build starts
4. Build fails
5. Fix the problem (maybe one or two files worth of fixes)
6. Run tup again
7. Scan starts over from scratch; waiting another few minutes.