[PATCH] D125946: Handles failing driver tests of clang

[Vassil Vassilev via Phabricator via cfe-commits]

v.g.vassilev added inline comments.


================
Comment at: clang/test/Interpreter/clangtests.cpp:1
+// RUN: clang-repl %S/../Lexer/badstring_in_if0.c -Xcc -E -Xcc -verify
+// RUN: clang-repl %S/../Lexer/unknown-char.c -Xcc -E -Xcc -verify
----------------
dblaikie wrote:
> dblaikie wrote:
> > v.g.vassilev wrote:
> > > @rsmith, would it be acceptable to have a test that refers other tests from the repo in that manner?
> > Generally that's not done - and the inputs should be moved into an "Inputs" subdirectory and shared from there. Tests that are in different subdirectories - not sure if there's a good way to share those, maybe with an "Inputs" directory in a parent directory of both tests? We might not have precednt for that
> But more broadly, could you explain what the goal of these tests are? Generally I would discourage what I think of as "shotgun" testing - taking some existing comprehensive test for a particular feature and using it to test a mostly orthogonal feature. The orthogonal feature should have more targeted tests/it shouldn't be using such broad testing in the regression suite here.
My take is that `clang-repl` is basically clang that takes inputs incrementally. Being that, means that we should be in a position to process whatever clang processes and thus we run against all of the existing tests. We planned to add the ones which we did not support as regression tests.

We can add more targeted tests but they would be copies or simplifications of already existing ones. Hence there is my hesitation - reuse or duplication...



CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D125946/new/

https://reviews.llvm.org/D125946