================
@@ -1,23 +1,10 @@
// RUN: rm -rf %t
// RUN: split-file %s %t
-//--- missing_tu.json.in
-[{
- "directory": "DIR",
- "command": "clang -fsyntax-only DIR/missing_tu.c",
- "file": "DIR/missing_tu.c"
-}]
-//--- missing_header.json.in
-[{
- "directory": "DIR",
- "command": "clang -fsyntax-only DIR/missing_header.c",
- "file": "DIR/missing_header.c"
-}]
//--- missing_header.c
#include "missing.h"
-// RUN: sed -e "s|DIR|%/t|g" %t/missing_tu.json.in > %t/missing_tu.json
-// RUN: not clang-scan-deps -compilation-database %t/missing_tu.json
2>%t/missing_tu.errs
+// RUN: not clang-scan-deps -- "clang" %t/missing_tu.c 2>%t/missing_tu.errs
----------------
jansvoboda11 wrote:
LIT complains about unquoted `clang` in attempt to nudge people towards using
`%clang`. That, however, adds a bunch of default flags to the invocation and
I'm not sure I want those in our scanner tests (especially when testing the
search path optimization for example).
The "-fsyntax-only" flag was removed because it's unnecessary (it was
originally used to prevent the scanner actually generating compilation output
files I think) and because it implies no output file (which the per-file mode
currently requires). Since the scanner is intended to be run on actual
compilation jobs, I think it's even desirable removing stuff like
"-fsyntax-only".
https://github.com/llvm/llvm-project/pull/88764
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
