[Lldb-commits] [PATCH] D13028: [RFC] Merge dsym and dwarf test cases
Zachary Turner via lldb-commits
lldb-commits at lists.llvm.org
Mon Sep 21 15:57:57 PDT 2015
This is great, I've been thinking about implementing the exact same thing
Is it possible to remove the need to call self.build()? It seems like the
the wrapped_run function could do it before invoking the actual test
method. This way a large majority of tests could be made simpler.
If you want to customize the build process, you could use a decorator like:
If no @build_properties decorator is present, it just calls self.build()
before invoking the test.
On Mon, Sep 21, 2015 at 1:46 PM Tamas Berghammer <tberghammer at google.com>
> tberghammer added a comment.
> I decided to make it an opt-out feature as most of the test case will use
> some debug info and this way you don't have to specify it explicitly
> (mutating a test what don't have any debug info only have performance hit
> because it will run the same test multiple times).
> About mutating, my current plan is to mutate every test isn't marked as
> no_debug_info_test and mark the test with dwarf_test / dsym_test / dwo_test
> annotations. The purpose of these annotation would be to enable the debug
> info based test filtering (e.g. the already existing "-N dwarf" option) and
> it should also test if the current system (compiler, OS) supports the given
> debug info format and skip the test if necessary (the dsym_test annotation
> already skip the test on non-darwin systems).
> Other option is to generate the mutated tests only on platforms where they
> are supported. It have the advantage of removing a huge number of skipped
> tests and reducing the noise with it, but it also makes it more difficult
> to compare platforms (and possibly mask away some basic coding errors). I
> don't have a strong opinion about the two options so if somebody have
> preferences then let me know.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the lldb-commits