Enable travis #7
Conversation
|
Nice! You're a miracle worker. Now it seems to fail on 1.8 only because I've used |
|
What prompted you to add the bang equivalents? |
|
It occurred to me that |
|
Sure. I can add tests for the new methods, no prob. I can also switch away from |
|
I'd rather remove 1.8.7 support, it's such a security risk! But if you're set on legacy support, feel free to fix the |
|
If we would be able to remove a bunch of C code if we dropped support for 1.8, then I would be up for it. But if it's the same to us either way, then let's keep it since the previous version had it. |
|
Nah, the code will remain the same. Go ahead and fix the tests then -- I think this is ready to work everywhere. |
Instead of using newer Array#sample, implement shuffling of an array manually.
I've picked the earliest test-unit version that doesn't define hoe as runtime dependency.
mislav
force-pushed
the
enable-travis
branch
from
44434dc
to
45cd9f5
Compare
It seems like recent C changes might have reduced the standard deviation
in `sort`, but not significantly, and it varied during runs:
VersionSorter .sort
Baseline: | X---------------|
Current: | X----- |
0 744.500 us
VersionSorter .rsort
Baseline: | X-------|
Current: | X------|
0 642.000 us
Recording the new baseline just in case.
Let's see if we can get this to build on as many platforms as possible.
cc @mislav