• Made the patch affect only -fdiagnostics-absolute-paths option.

• Changed the format to <archive name>(<object name> at <offset>);
• Changed "if" to "?:";
• Simplified a check in the tests.

• Updated according to @MaskRay's comments.

In D60964#1485943, @jhenderson wrote:

LGTM, I think (not verified the regex though).

• Simplified print statements.
• Added escaping in the regex string.

Thanks!

• Rebased on top of D60816; updated the documentation accordingly.

• Extended an example for the --output-style option.

• Added --implicit-check-not=main and --implicit-check-not=inctwo to the checks in output-style-inlined.test. I added the similar option to the other two checks so that all the checks in the test follow the same pattern.

In D60067#1468404, @jhenderson wrote:

Code all looks good, but you should talk about the effect of --output-style on the output for both llvm-symbolizer and llvm-addr2line as appropriate.

• Simplified tests.

• Extracted the changes for --output-style into a separate patch, D60770.

It looks like the changes for --output-style should go into a separate patch with the corresponding tests.

OK, let be it.

• rst -> md.
• Made printing of an empty line depend on the -output-style option instead of the tool name.

• Fixed a typo.

• Added a doc page for llvm-addr2line. I'll really appreciate any corrections.
• Addressed other review comments.

• Rebased the patch to the current tip;
• Moved the code which handles differences in printing source positions in llvm-symbolizer.cpp;
• Reworked tests.

Please note, that, in contrary to POSIX OSes, on Windows a path like dir1\<junction point>\.. refers to dir1. That's why we do not need to ask the system for the fully expanded path and can do all calculations internally.

It looks like I've just forgotten to enable a prefix form for the switch. Does that resolve the doubts?

In D60196#1453050, @jhenderson wrote:

Do you think it's worth testing the behaviour of -e in the middle of a group of options e.g. "-Cefi"?

Thanks! I believe you are right. I'll put explicit addresses on commit.

LGTM.

By the way, we have some inconsistency in how cl::Prefix and cl::AlwaysPrefix options work in case of -opt=val form. The former takes only val, while the latter keeps =. Moreover, prefix grouping options in GNU's binutils, like -j in objdump, work partially as ours cl::AlwaysPrefix, because they preserve =, and partially as cl::Prefix, because they allow passing the value in a separate argument. Thus, we still cannot fully reproduce that behavior.

• Fixed grammar and formatting issues.
• Added test cases to check prefix options when their values are textually equal to another option.

In D58499#1408521, @jhenderson wrote:

LGTM.

Thanks!

In D58499#1405561, @jhenderson wrote:

What happens for ValueOptional? I want to make sure that it doesn't treat the other grouped letters after it as the option's value.

• Updated the documentation (feel free to correct my wording).
• Amended a comment in the test.

For me, the code should be, above all, reliable. You can trust it only in that case; otherwise, you never know, when and how it is going to be broken and produce an unexpected result.

Thanks! I'll wait a bit, but not very long. I believe the patch is safe because the default behavior is not changed.

Updated the test.

• Made DefaultAltIdx private.
• Extended the test.
• Addressed other review comments.

Ping.

Addressed review comments.