Introduce a toolset getrelative toolset function to allow customization.
#2244
Conversation
|
Should also apply to other toolsets (so currently dotnet and msc, as others use |
This just centralizes all `getrelative`, allowing for customization by external code, by overriding this function. This is needed for instance in Ninja generators, where the location of a project is not the workspace location, but the code of the Ninja file itself should still refer to workspace-relative paths.
tritao
force-pushed
the
gcc-toolset-getrelative
branch
from
37d1cd0
to
6d817ad
Compare
|
Updated the |
tritao
changed the title
gcc.getrelative toolset function to allow customization.getrelative toolset function to allow customization.
tritao
force-pushed
the
gcc-toolset-getrelative
branch
from
e0868af
to
2ad4923
Compare
|
Updated this again with some more thorough testing to at least make sure these exist and continue to exist for all toolsets. @nickclark2016 Please take another look when you have a chance. |
|
Will leave open for comment for a few days. Barring any change requests, will merge. Thank you! |
| prepare() | ||
| local toolset = p.config.toolset(cfg) | ||
| print(toolset) | ||
| test.isnotnil(toolset.getrelative) |
There was a problem hiding this comment.
Your test is really unit,
and I think it doesn't test what you want to check, i.e clang.some_function_which_uses_getrelative() calls actually clang.getrelative (especially with gcc redirection)
And I think that test would currently fail :-(
There was a problem hiding this comment.
Can you clarify what you mean by that?
There was a problem hiding this comment.
I meant, if user (or premake.ninja) does clang.getrelative = function (cfg, path) return p.workspace.getrelative(cfg.workspace, value) end
then clang.getincludedirs(cfg.includedirs, ..) would call gcc.getincludedirs(cfg.includedirs, ..) which will call
gcc.getrelative and not clang.getrelative.
As bug happens with premake.ninja, can you test your changes with problematic case (project location != workspace location) with clang toolset? (https://github.com/Jarod42/premake-sample-projects/tree/location/projects/project-location is one such project (not the main branch))
There was a problem hiding this comment.
Yeah thats an issue.
What do you think, should we forgo this approach and instead have a global overrideable p.tools.getrelative?
There was a problem hiding this comment.
A priori, I don't see issue with p.tools.getrelative
This just centralizes all
project.getrelativein the GCC toolset, allowing for customization by external code, by overriding this function.This is needed for instance in Ninja generators, when the location of a project is not the workspace location, but the code of the Ninja file itself should still refer to workspace-relative paths.