superflore-gen-oe-recipes: Delete all generated files when regenerating all recipes

[shr-project]

• the overlay.repo.remove_file(existing, True) call in bitbake/genpackages.py works fine, but only when the new recipe is being generated in the same directory as the previous version: existing = glob.glob('{}*.bb'.format(prefix)) where prefix is: prefix = '{0}/meta-ros{1}-{2}/generated-recipes/{3}/{4}' which will result in multiple recipes for the same package when the package is moved to different prefix like pcl-conversions did for dashing and eloquent:

  meta-ros$ find . -name pcl-conversions\*bb
  ./meta-ros2-eloquent/generated-recipes/perception-pcl/pcl-conversions_2.1.0-1.bb
  ./meta-ros2-eloquent/generated-recipes/pcl-conversions/pcl-conversions_2.0.0-1.bb
  ./meta-ros1-melodic/generated-recipes/perception-pcl/pcl-conversions_1.7.0-2.bb
  ./meta-ros2-crystal/generated-recipes/pcl-conversions/pcl-conversions_2.0.0.bb
  ./meta-ros2-dashing/generated-recipes/perception-pcl/pcl-conversions_2.0.0-1.bb
  ./meta-ros2-dashing/generated-recipes/pcl-conversions/pcl-conversions_2.0.0-1.bb

  and especially for dashing where the version is identical it's undeterministic which recipe bitbake will use. And the recipe wasn't just moved to different directory, it's quite different:

  diff ./meta-ros2-dashing/generated-recipes/perception-pcl/pcl-conversions_2.0.0-1.bb ./meta-ros2-dashing/generated-recipes/pcl-conversions/pcl-conversions_2.0.0-1.bb
  9c9
  < AUTHOR = "Paul Bovbel <paul@bovbel.com>"
  ---
  > AUTHOR = "Chris Lalancette <clalancette@openrobotics.org>"
  14c14
  < LIC_FILES_CHKSUM = "file://package.xml;beginline=16;endline=16;md5=d566ef916e9dedc494f5f793a6690ba5"
  ---
  > LIC_FILES_CHKSUM = "file://package.xml;beginline=12;endline=12;md5=d566ef916e9dedc494f5f793a6690ba5"
  16c16
  < ROS_CN = "perception_pcl"
  ---
  > ROS_CN = "pcl_conversions"
  19a20
  >     builtin-interfaces \
  21d21
  <     message-filters \
  23,24d22
  <     pcl-msgs \
  <     rclcpp \
  33a32
  >     builtin-interfaces \
  35d33
  <     message-filters \
  37,38d34
  <     pcl-msgs \
  <     rclcpp \
  46,47d41
  <     libeigen \
  <     message-filters \
  49,52d42
  <     pcl-msgs \
  <     rclcpp \
  <     sensor-msgs \
  <     std-msgs \
  67c57
  < # matches with: https://github.com/ros2-gbp/perception_pcl-release/archive/release/dashing/pcl_conversions/2.0.0-1.tar.gz
  ---
  > # matches with: https://github.com/ros2-gbp/pcl_conversions-release/archive/release/dashing/pcl_conversions/2.0.0-1.tar.gz
  69,70c59,60
  < SRC_URI = "git://github.com/ros2-gbp/perception_pcl-release;${ROS_BRANCH};protocol=https"
  < SRCREV = "0f550985da8bd73fd7451350c63e1ce96e78c955"
  ---
  > SRC_URI = "git://github.com/ros2-gbp/pcl_conversions-release;${ROS_BRANCH};protocol=https"
  > SRCREV = "8d0d0ef7d5d253f11fbf52388d64fad59e46377a"

• grep in the current build logs shows that the "older" meta-ros2-dashing/generated-recipes/pcl-conversions/pcl-conversions_2.0.0-1.bb was currently being used in dashing instead of "newer" meta-ros2-dashing/generated-recipes/perception-pcl/pcl-conversions_2.0.0-1.bb

• corresponding rosdistro change for dashing: https://github.com/ros/rosdistro/commit/19b06c44df69f9bf44dd7ab1b74236e9e386c848 and eloquent: https://github.com/ros/rosdistro/commit/67c3ef53979ea8d85254bdfc427b142eda69fb57

Signed-off-by: Martin Jansa martin.jansa@lge.com

[shr-project]

Drop this and what follows. I don't see how showing how different the recipes are helps further explain the need for the change.

I disagree, It shows why it's important to implement this change.

[herb-kuta-lge]

I disagree, It shows why it's important to implement this change.

Wouldn't still be important to implement it even if the recipes were identical?

[shr-project]

Wouldn't still be important to implement it even if the recipes were identical?

No it wouldn't be important, because bitbake doesn't care from which directory the recipe was used, but when the built output is very different, then it's important to use the new recipe.

[herb-kuta-lge]

even if the recipes were identical

Actually, if a recipe moves to a different subdirectory, it's ROS_CN setting will be different, so the recipes will never be identical.

[herb-kuta-lge]

I don't see how showing how different the recipes

I disagree, It shows why it's important to implement this change.

The details of the differences are unimportant, so perhaps show how different they are by piping the diff output into diffstat? (Use diff -C 1 so that the filename appears in the diffstat output.)

[shr-project]

Comments explaining what would cause unexpected output from git status --porcelain and glob.

Can you please provide these comments?

I don't know what unexpected might have happen especially in combination with --only which isn't implemented yet AFAIK.

[herb-kuta-lge]

I don't know what unexpected might have happen especially in combination with --only which isn't implemented yet AFAIK.

If it's unexpected, shouldn't it be considered an ASSERT that stops execution?

[shr-project]

The output is verbose, but err() output is quite rare in it and contains even more important messages than this unexpected output, see the errors in last meta-ros regeneration: http://paste.ubuntu.com/p/b6JbJyTp8d/

[herb-kuta-lge]

OK, then let's leave it as is.

[herb-kuta-lge]

@allenh1 < Now that this last open PR was been merged, could you release a v0.3.2 ?

[allenh1]

@herb-kuta-lge That sounds like a good idea to me!

@tfoote FYI