switch to tf2 #31
switch to tf2 #31
Conversation
|
Looks fine to me, but @tfoote should really review it, he's more familiar with the tf specific changes than me. |
|
Reading through it it looks reasonable. I opened an enhancement ticket to duplicate the getYaw method from tf. But you shouldn't wait for that. |
|
thx ! With all the tf2 conversions I've done, the only thing that could not be done simply is indeed the getYaw function (navigation uses it a lot). Will follow up on ros/geometry2#110 |
vrabaud
force-pushed
the
hydro-devel
branch
3 times, most recently
from
33dc6de
to
c6cea78
Compare
vrabaud
force-pushed
the
hydro-devel
branch
2 times, most recently
from
88c76b2
to
3c5cdc7
Compare
|
@vrabaud Is there any reason that this didn't ever get merged? |
|
@tfoote, can you please review ? Also, tests now run slower, is that to be expected from tf2 ? Thx |
gmapping/src/tftest.cpp
Outdated
| tf_ = new tf::TransformListener(*node_, true, | ||
| ros::Duration(10)); | ||
| tf_.reset(new tf2::Buffer(ros::Duration(10))); | ||
| tfl_ = new tf2_ros::TransformListener(*tf_, *node_, true); | ||
| tf_->setExtrapolationLimit( ros::Duration().fromSec(0.2)); |
There was a problem hiding this comment.
setExtrapolationLimit was chosen not to be implemented in tf2 as there wasn't yet found a good use case for it in tf. I'm not sure that this unit will compile.
There was a problem hiding this comment.
good catch, that file has been useless forever actually.
| @@ -4,5 +4,5 @@ | |||
| <!--test time-limit="30" test-name="basic_localization_stage" pkg="gmapping" | |||
| type="test_map.py" args="90.0"/--> | |||
| <!-- TODO: Fix the rtest to work with replay, for now it's not working --> | |||
| <test time-limit="250" test-name="map_data_test_replay" pkg="gmapping" type="gmapping-rtest" args="200.0 0.05 4000 4000 0.005 0.010"/> | |||
| <test time-limit="300" test-name="map_data_test_replay" pkg="gmapping" type="gmapping-rtest" args="200.0 0.05 4000 4000 0.005 0.010"/> | |||
There was a problem hiding this comment.
The changes to tf -> tf2 should not change the behavior. tf is using tf2 under the hood these days so if anything it should be slightly faster due to fewer function calls due to no indirection. This suggests that there's some timeouts or delays coming in somewhere.
There was a problem hiding this comment.
performance was actually very close to that limit (I have it to fail sometimes at that limit before this patch).
There was a problem hiding this comment.
Yeah, you don't want to be right on the cusp anyway, that a forumla for flaky tests.
@tfoote , @wjwwood , if you have a few minutes to check that. Tests pass. Thx