- User Since
- May 9 2013, 11:10 AM (253 w, 2 d)
Thu, Mar 8
Added more commentary. Now back to round 3 of getting the "file 0" patch in...
Wed, Mar 7
Tue, Mar 6
So non-verbose shows the section-relative offset instead of the CU-relative offset. That seems reasonable.
You might give Adrian and David another day to complain, but LGTM.
Mon, Mar 5
+ @eugenis as this seems like his area.
Fri, Mar 2
Wed, Feb 28
Tue, Feb 27
LGTM with minor changes.
I didn't try to verify that you found all the places where this check would be needed; I trust that you will trip over them if you missed any. :-)
Mon, Feb 26
Fri, Feb 23
Being a cross-compiler I think it's generally a good thing to have more combinations be less broken.
Note that PS4's compiler is hosted on Windows and uses the gcc-style driver; it's convenient sometimes to be able to target Windows without having to learn a new driver language. So thanks for doing this!
Looks generally straightforward. I'd give other people a chance to chime in but I have only minor comments.
It seems a little counter-intuitive for embedded source to have a filename, but I guess that's so the debugger has a familiar-looking handle to refer to it?
Thu, Feb 22
Do you have a use-case for the C API? We tend to update it only when necessary.
And when we do update it, it has to be backward-compatible IIUC. Adding new parameters to an existing API does not preserve that property.
Wed, Feb 21
I'm not really familiar with dsymutil's use cases; does it operate on .o files? If so, maybe you can generate a test with gcc -g -S and reduce the resulting assembler by hand? If the assembler isn't ridiculously long.
Tue, Feb 20
Fri, Feb 16
Use Expected<unsigned> instead of just unsigned to return errors.
Fix the temp-file thing in the test and LGTM.
Actually never mind. Losing all the debug info would also pass that check, so my previous 3-line recommendation stands.
Instead of all the elaborate checks, you could use llvm-dwarfdump -debug-str (dump just the .debug_str section) and then use CHECK-NOT: " to prove it's empty.
Feb 14 2018
Feb 12 2018
Please make sure to cite the PR in the commit message.
Feb 9 2018
I'd have though invalid.s should be a text file from the name, but it shows up as binary here?
Feb 8 2018
Feb 7 2018
Feb 6 2018
Feb 5 2018
Address review comments:
Pass the MCDwarfLineStr around by pointer, eliminating the operator-bool;
use new dwarfdump feature in the test to show the section for the string references;
put more responsibility on the (only) user of the class.
Feb 1 2018
+1 for testcase.
@JDevlieghere are the quoted file/directory names compatible (enough) with Darwin's dwarfdump? I know you were working to make them compatible (although maybe it was mostly command-line compatible).
Address review comments.
WFM. The practical description of layering will be enough for people trying to write patches. The software-engineering-process geek in me would still prefer a broader description but he won't make any more fuss about it. :-)
Header looks fine.
Jan 31 2018
Jan 30 2018
Jan 29 2018
Jan 26 2018
If you want to force DWARF 2, probably clamping the version in LLVM would be simpler? Although most of the debug-info tests are architecture-specific and wouldn't run for an NVPTX target anyway.
I'm okay with this, but as I said I am not so familiar with the metadata API and I think someone else should give the final approval.
And thanks for taking out the mis-handling of DW_AT_discr_value in the dumper!
Jan 25 2018
Most of this is just plumbing through cchanges to track the new section. DWARFFormValue.cpp is really the missing functional bit for dumping purposes.