We currently include the AVM2 playerglobals abc in the repo, but not the AVM1 one. We include the checksums file for neither. Both should be changed so botio doesn't spend several minutes building abc files for which we know with certainty that they don't need rebuilding. Yury, is there any problem with this, or is it just a matter of changing .gitignore files and committing?

(In reply to Till Schneidereit [:till] from comment #0) > We currently include the AVM2 playerglobals abc in the repo, but not the > AVM1 one. We don't include playerglobals abc, so we have to build it as well. We include avm1lib though (similar to builtin.abc and shell.abc), not sure why we spend time to rebuild that we are already including that. > We include the checksums file for neither. Both should be changed > so botio doesn't spend several minutes building abc files for which we know > with certainty that they don't need rebuilding. Ideally, botio must rebuild all binary file before testing (including avm1lib.abc and builtin.abc) to avoid disconnect between source and binaries. The botio shall just play a role of the gate before landing. It does not make any sense to trigger botio for your own testing.

(In reply to Yury Delendik (:yury) from comment #1) > We don't include playerglobals abc, so we have to build it as well. We > include avm1lib though (similar to builtin.abc and shell.abc), not sure why > we spend time to rebuild that we are already including that. Oh, right, we don't include playerglobals.abc, but the others. > > > We include the checksums file for neither. Both should be changed > > so botio doesn't spend several minutes building abc files for which we know > > with certainty that they don't need rebuilding. > > Ideally, botio must rebuild all binary file before testing (including > avm1lib.abc and builtin.abc) to avoid disconnect between source and > binaries. That's debatable. The checksum system guarantees that there is no divergence, so as long as we have the binaries comitted, we should also commit the checksums to not waste compilation time for nothing. > The botio shall just play a role of the gate before landing. It > does not make any sense to trigger botio for your own testing. Obviously. I want to be able to land things fairly quickly, though, and botio is slow. Maybe before doing anything else we should move it to faster hardware, but the checksums thing would be a simple enough change.

Assigning to Till because, IIUC, Yury says only Till cares about botio perf. :)

We don't have the AVM2 abc anymore, and will get rid of the avm2 one, too, so this won't happen.