New option format for prefix mapping#10723
Conversation
|
@swift-ci Please test LLVM |
benlangmuir
left a comment
benlangmuir
left a comment
There was a problem hiding this comment.
We should have a new Driver test that checks the new spelling and confirms the ordering when you mix it with the old spelling is preserved in the cc1 command line.
| } | ||
| } | ||
|
|
||
| for (const Arg *A : Args.filtered(options::OPT_fdepscan_prefix_map)) { |
There was a problem hiding this comment.
We should make a single loop that handles both OPT_fdepscan_prefix_map and OPT_fdepscan_prefix_map_EQ so that the order is preserved with either spelling.
There was a problem hiding this comment.
I implemented this in my recent commit, could you please check that? 😄
cachemeifyoucan
left a comment
cachemeifyoucan
left a comment
There was a problem hiding this comment.
Current version looks good.
Can you add some lit tests to use the new flag you added and also mixing it together with the old flag?
sina-mahdavi
force-pushed
the
sina-mahdavi/prefix-mapping-fix
branch
from
e5bb87c to
40f8661
Compare
|
There are some |
sina-mahdavi
force-pushed
the
sina-mahdavi/prefix-mapping-fix
branch
from
40f8661 to
83b5c98
Compare
benlangmuir
left a comment
benlangmuir
left a comment
There was a problem hiding this comment.
The implementation LGTM, so I think we just need a new driver test that checks the behaviour of mixing this with the old flag.
|
@swift-ci please test llvm |
benlangmuir
left a comment
benlangmuir
left a comment
There was a problem hiding this comment.
LGTM assuming PR tests don't show any related failures. Thanks!
I think there's some tests failing in Edit: should be fixed now, |
|
@swift-ci please test llvm |
|
Alright, I think the tests are okay now. The only failing tests are in |
…to... ...`clang::tooling::dependencies::DepscanPrefixMapping::configurePrefixMapper` See swiftlang/llvm-project#10723
4 participants
The current option for specifying prefix mappings
-fdepscan-prefix-map=<old>=<new>fails to handle paths containing=characters properly. This PR adds a 2nd form of that option-fdepscan-prefix-map old newwhile keeping the old format as well to fix this problem.