I filed https://bugs.llvm.org/show_bug.cgi?id=40229 and changed who is CCed on this bug.

The bug seems that vpx_subpixel_8t_ssse3.o has read only ".rodata" section, but embedded.S/embedded.o generated in V8 has read/write ".rodata" section. MSVC link.exe complains this mismatch when merging ".rodata" section, but it seems lld-link.exe doesn't do this validation.

Thanks for finding that Tom. How did you identify which .o file was conflicting?

Having a .rodata file tagged as read/write seems bad. Quite bad.

Adding a v8 owner who has worked on embedded.o for comment.

I set break point in link.exe when creating a new ".rodata" section, and found the r/w one comes from embedded.S.

To repro it, you can assemble below assembly like "clang-cl -c a.S" then run "link /dump /headers a.obj" to check the property of ".rodata" section.

.section .rodata
.long 1

which shows below output:
SECTION HEADER #4
 .rodata name
       0 physical address
       0 virtual address
       4 size of raw data
      B4 file pointer to raw data (000000B4 to 000000B7)
       0 file pointer to relocation table
       0 file pointer to line numbers
       0 number of relocations
       0 number of line numbers
C0100040 flags
         Initialized Data
         1 byte align
         Read Write

Below is the proposed patch:

https://chromium-review.googlesource.com/c/v8/v8/+/1396361

Thanks for investigating, interesting indeed. I'll take this onto my queue but Tom's fix from #5 lgtm.

The following revision refers to this bug:
  https://chromium.googlesource.com/v8/v8.git/+/934af8dde90daed8b211df0c724a3053c701e9e4

commit 934af8dde90daed8b211df0c724a3053c701e9e4
Author: Tom Tan <Tom.Tan@microsoft.com>
Date: Wed Jan 09 07:58:51 2019

.rodata from embedded.S should be read only

.rodata usually hosts read only data. MSVC link.exe complains mismatch when
merging this read/write .rodata from embedded.S with .rodata from other object
file.

Bug:  chromium:919180 
Change-Id: I7789e42afe116cc4bf772e2cbb312d19e4ce7fe5
Reviewed-on: https://chromium-review.googlesource.com/c/1396361
Commit-Queue: Jakob Gruber <jgruber@chromium.org>
Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Cr-Commit-Position: refs/heads/master@{#58653}
[modify] https://crrev.com/934af8dde90daed8b211df0c724a3053c701e9e4/src/snapshot/embedded-file-writer.cc
[modify] https://crrev.com/934af8dde90daed8b211df0c724a3053c701e9e4/tools/snapshot/asm_to_inline_asm.py

The bug is marked as P3 or Feature. It should not be merged as M72 is in beta. 
Please contact the approriate milestone owner if you have questions.
Owners: govind@(Android), kariahda@(iOS), djmm@(ChromeOS), abdulsyed@(Desktop)

For more details visit https://www.chromium.org/issue-tracking/autotriage - Your friendly Sheriffbot

Updated labels, reapplying merge request. 

The CL in #8 fixes an issue that causes a section of the binary to be read-write that is intended to be read-only.

This bug requires manual review: M72 has already been promoted to the beta branch, so this requires manual review
Please contact the milestone owner if you have questions.
Owners: govind@(Android), kariahda@(iOS), djmm@(ChromeOS), abdulsyed@(Desktop)

For more details visit https://www.chromium.org/issue-tracking/autotriage - Your friendly Sheriffbot

brucedawson@ are you comfortable with fix in #8 for M72 branch?

The fix in #8 seems very safe, no concerns here.

There's currently a CL in flight with a better, simpler fix: https://crrev.com/c/1407240. 

The following revision refers to this bug:
  https://chromium.googlesource.com/v8/v8.git/+/485136287e377f1b6bd8205d215f872b1db20629

commit 485136287e377f1b6bd8205d215f872b1db20629
Author: Tom Tan <Tom.Tan@microsoft.com>
Date: Mon Jan 14 08:15:11 2019

Use .rdata as section name for asm targeting COFF

.rdata is the default section which hosts read-only data for COFF. Use this
default section name avoids creating a new .rodata section with explicit
read-only property.

Bug:  chromium:919180 
Change-Id: I7325cbcfdb142b3ee15de93b7881f755c365d6e6
Reviewed-on: https://chromium-review.googlesource.com/c/1407240
Commit-Queue: Jakob Gruber <jgruber@chromium.org>
Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Cr-Commit-Position: refs/heads/master@{#58764}
[modify] https://crrev.com/485136287e377f1b6bd8205d215f872b1db20629/src/snapshot/embedded-file-writer.cc
[modify] https://crrev.com/485136287e377f1b6bd8205d215f872b1db20629/tools/snapshot/asm_to_inline_asm.py

great thanks! brucedawson@ - are you comfortable with #16?

Yep, seems good. The new fix works well.

Approving #16 for M72, branch:3626

Backmerge in flight on top of refs/branch-heads/7.2: https://crrev.com/c/1411887

The following revision refers to this bug:
  https://chromium.googlesource.com/v8/v8.git/+/5047389f9ab470322a6acf30da4daad505e1da1a

commit 5047389f9ab470322a6acf30da4daad505e1da1a
Author: Jakob Gruber <jgruber@chromium.org>
Date: Thu Jan 17 07:28:34 2019

Merged: Use .rdata as section name for asm targeting COFF

.rdata is the default section which hosts read-only data for COFF. Use this
default section name avoids creating a new .rodata section with explicit
read-only property.

No-Try: true
No-Presubmit: true
No-Treechecks: true
Tbr: hablich@chromium.org
Bug:  chromium:919180 
Change-Id: I7325cbcfdb142b3ee15de93b7881f755c365d6e6
Reviewed-on: https://chromium-review.googlesource.com/c/1407240
Commit-Queue: Jakob Gruber <jgruber@chromium.org>
Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Cr-Original-Commit-Position: refs/heads/master@{#58764}
Reviewed-on: https://chromium-review.googlesource.com/c/1411887
Cr-Commit-Position: refs/branch-heads/7.2@{#41}
Cr-Branched-From: 6acd03c9b8a8232aee95f25fbf6ae822aaedae75-refs/heads/7.2.502@{#1}
Cr-Branched-From: b03041de094610ef24e0e4fb6bf4c700fa1553ed-refs/heads/master@{#57910}
[modify] https://crrev.com/5047389f9ab470322a6acf30da4daad505e1da1a/src/snapshot/embedded-file-writer.cc

Bruce, assigning back to you in case you want to leave this open - feel free to close otherwise, backmerges are done.

Closing as fixed. The llvm bug will track getting lld-link to notice this situation.