#1479 closed task (fixed)
Makefile/setup.py cleanups
| Reported by: | warner | Owned by: | somebody |
|---|---|---|---|
| Priority: | major | Milestone: | 1.9.0 |
| Component: | code | Version: | 1.8.2 |
| Keywords: | Cc: | ||
| Launchpad Bug: |
Description
Ticket #1454 had a non-debian patch in attachment:simplify-makefile-and-setup-py.darcs.patch:ticket:1454 which needs separate review.
Change History (6)
comment:1 follow-up: ↓ 3 Changed at 2011-08-11T18:52:18Z by warner
comment:2 Changed at 2011-08-11T18:58:43Z by warner
- Resolution set to fixed
- Status changed from new to closed
Landed, in 16e6bb2b738e4edc.
comment:3 in reply to: ↑ 1 Changed at 2011-08-12T00:10:41Z by davidsarah
Replying to warner:
Replacing $(RUNPP) with something in our bin/tahoe loses the ability to get a repl running without actually importing the tahoe code (at least allmydata.scripts and that infernal __init__.py).
Well, this guarantees that we do the same checks on package versions (which are necessary to detect the situation in #1258) when starting a repl as we do when running the CLI, and for the unit tests (#1296). The path setup code is so flaky that I really prefer that there is only one kind of flakiness rather than several.
Relying more heavily upon the gross 'darcsver'-prefixing in setup.cfg is bad, since I want to get rid of setuptools and this reliance makes that more difficult, but I don't have the energy to fight for the Makefile any more.
When we get rid of the prefixing in setup.cfg, it'll automatically be got rid of for the Makefile targets as well.
So I'm +1 on this as long as we really try to make bin/tahoe and __init__.py and our setup.py simpler in the near future.
Just try and stop me! ;-)
comment:4 Changed at 2011-08-12T00:11:24Z by davidsarah
- Keywords needs-review removed
comment:5 Changed at 2011-08-26T23:05:05Z by david-sarah@…
In bdca62ea02097f93:
comment:6 Changed at 2011-08-26T23:07:15Z by davidsarah
I had deleted test_mac_diskimage from setup.py but not from setup.cfg; bdca62ea02097f93 fixes that.
Replacing $(RUNPP) with something in our bin/tahoe loses the ability to get a repl running without actually importing the tahoe code (at least allmydata.scripts and that infernal __init__.py). I don't think that's a big loss, though. I *do* like the removal of code from setup.py.
Relying more heavily upon the gross 'darcsver'-prefixing in setup.cfg is bad, since I want to get rid of setuptools and this reliance makes that more difficult, but I don't have the energy to fight for the Makefile any more. So I'm willing to go along with that too.
So I'm +1 on this as long as we really try to make bin/tahoe and __init__.py and our setup.py simpler in the near future.