twoh added inline comments.
================
Comment at: test/Frontend/preprocessed-input.c:3
+// RUN: %clang -emit-llvm -S -o - %t.i | FileCheck %s
+// CHECK: source_filename = {{.*}}preprocessed-input.c"{{$}}
----------------
inglorion wrote:
> Actually, I think you don't even have to run the preprocessor - you can just
> put the file with the linemarker here and check that the name from the
> linemarker is propagated, right?
>
> Also, it seems that there is a similar issue to the one you're trying to fix
> when going from .ll to .o (or .ll to .s, for that matter) - the name is taken
> from the file you're reading from, not from the source_filename directive in
> that file. Of course, that doesn't differ from gcc (given that gcc doesn't
> handle .ll files), but you may want to address that, too, for consistency.
@inglorion Thanks for the comments! Putting preprocessed file makes sense to
me.
I recognize the second issue. If I do something like
```
clang -E -o test.i /abspath/test.c
clang -g -c -o test.o test.i
```
DW_AT_name is still test.i (for gcc, it is taken from the linemarker as well).
Moreover, if we use the absolute path of test.i for clang, i.e.
```
clang -g -c -o test.o /abspath/test.i
```
DW_AT_name becomes /abspath/test.i where DW_AT_comp_dir is /abspath. I think
this could be a problem if the user of the object file assumes that DW_AT_name
is a relative to DW_AT_comp_dir. I'm planning to address this with a separate
patch.
https://reviews.llvm.org/D30663
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
