Hi Douglas,

Hi Douglas,

@StephenTozer Broken greendragon tests temporarily XFAILed with 37a20cc68f545647e614c5ba4ae311dc3fd277e9

LGTM! Thanks Alexey!

Thanks for the review Alexey.

All dwarf 5 cus get their strings converted to strxs.

Fixed comment for end offset

Landed with ab7874e945f0ecba4a62605950503b4c5d830c0b

Removed AddrPool.empty() check from cloneAllCompileUnits

Rebased on top of https://reviews.llvm.org/D155666 and https://reviews.llvm.org/D154638

Hi @avl Thanks for all the reviews, if you could give me the okay for this patch, I can push the whole stack. Thank you!

@avl I think those pre-integration errors are happening because phabricator doesn't have the binaries

Updated verify.test to fail in a different way, as suggested by Jonas and Alexey

@avl Hi Alexey, can you please let me know if this patch looks okay now? Thanks

Awesome! Thanks @JDevlieghere and @avl I will change verify.test to make sure there is something in there dsymutil cannot fixup.

Rebased on top of main

Rebased on top of main

@avl I imagine the patch application failed because this patch depends on https://reviews.llvm.org/D154638 going in?

Rebased on top of main

Fix test failures in dwarf5-loclists.test

Return proper size of DW_FORM_addrx, fix dwarf5-addresses.test

Pass AddrPool to DWARFStreamer::emitDwarfDebugLocListFragment` to calculate baseaddrx index there instead of passing index in

Right. We need to clean(), sorry. I missed the fact that DIECloner enumerates several compile units.

Does the rest of the patch look okay?

rebase on top of https://reviews.llvm.org/D154638

Add a test

Clear AddrPool after processing a compile unit

We have to clear the AddrPool after one compile unit is processed. If we consider the following test case with the attached files:

Rebased on top of https://reviews.llvm.org/D154638

Added new line at the end of dwarf5-addr_base.test

Always emit the form DW_FORM_addrx in the debug_info section when using dsymutil

Looks like @avl is correct:
Using just addrx, the size of the debug_info + the debug_addr + debug_abbrev sections is 903,287,425 Bytes or ~903 MB
Using addrx1, addrx2, addrx3, addrx4 the size of the debug_info + debug_addr + debug_abbrev sections is 905,953,864 MB or ~906 MB

I am not sure how to test this, how do I modify the compile unit in my test to not have a DW_AT_addr_base in it?

Fixed some comments

In D153734#4501034, @rastogishubham wrote:

I am reverting both fb087c17c8230 and 7f1028e9df52b to see if the bot becomes green again. I apologize for any incovenience

I am reverting both fb087c17c8230 and 7f1028e9df52b to see if the bot becomes green again. I apologize for any incovenience

The unit test LocateModuleCallbackTest.cpp is very flaky and is failing on the lldb arm64 incremental bot

Addressed review comments

1. Changed assert from UINT16_MAX + UINT8_MAX to UINT32_MAX >> 8

I am reverting the change for now, I apologize for any inconvenience

It looks like this change is breaking some lldb tests. https://green.lab.llvm.org/green/view/LLDB/job/lldb-cmake/57586/consoleFull#-1731262255a1ca8a51-895e-46c6-af87-ce24fa4cd561

Added assert to make sure the index value fits within the give addrx form

Patched the addr_base value, moved functions for emitting Addr section header, footer, and the section to be together, and used the same naming scheme as the other sections, and resolved nitpicks

Addressed review comments:

@avl

How about changing DW_RLE_start_length/DW_LLE_start_length into DW_RLE_startx_length/DW_LLE_startx_length for rnglists/loclists? Or this is for another patch?

In D154638#4478628, @avl wrote:

Could you compare size of .debug_info, .debug_addr sections for clang binary with and without this patch, please?

LGTM!

Added REQUIRES line to make test pass on bots that do not have the arm backend enabled

Trimmed down directory path