The change looks good to me; the other arches apparently use either bool or ParseStatus here.

• Rename getDispOpValue() to getImmOpValue().
• Rename FK_390_12 to FK_390_U12Imm and FK_390_20 to FK_390_S20Imm.
• Reformat.

I've added pc-relative tests, but only for PC32. For PC32DBL I tried (src-.)/2, but the common code does not recognize division as a relocatable expression, and does not let the backend to handle it either. GAS does not handle it either though, so it's probably fine:

• Emit relocations.
• Add various tests.
• Remove "imm" markup for expressions.
• Move the isExpr() logic to getDispOpValue().

• Avoid emitting empty cases entirely.

https://reviews.llvm.org/D155805

Thanks for letting me know, I'll fix this.

• s/MachineInstr/MCInst/g

• Document the new function.

• Move the endianness handling to the target C++ code.

• Fix the handling of multi-lit instructions on little-endian. The code used to return the start of the last lit, which is not useful.

• Pass endianness explicitly.

Right now there doesn't seem to be a generic way to get the instruction endianness.
The best example is ARM:

Hmm, yes, this indeed does not work for RISC-V. E.g., for LUI I get:

• Add little-endian support.
• Print operand names.
• Fix a typo in the #endif comment.

• Drop the ADDR64/GR64 patch.
• Adjust the tests to accept 0 instead of %r0 instead.
• There already are a lot of assembler tests that check that %r0 is accepted for ADDR64.

• Rebase (the dependency CL was merged).
• New prep patch: ADDR64/GR64 cleanup.
• Move InstRSEa.
• Improve whitespace.

Fortunately not. I found this while adding named sub-operand support to SystemZ.

This depends on https://reviews.llvm.org/D155158.

• Remove an unused variable.
• Remove an irrelevant whitespace change.

• Fixed an issue with copyRegSaveArea() overwriting shadow of caller's locals. The problem was that the kernel is compiled with "packed-stack", so register save areas may be smaller than 160 bytes. Since the kernel is compiled with "use-soft-float", it's enough (and also correct) to copy only 56 bytes. Now all kernel selftests pass.
• Added a test for this case.

• (unsigned)0 -> 0u (0 doesn't work, because the overload becomes ambiguous).
• Improve matching in the testcase.

• Add a comment to getOrInsertMsanMetadataFunction().

• Better explain the ABI situation.
• Extend the test.

The kernel patch is currently WIP; I'm trying to get rid of false positives.
Just some examples of the findings so far:

I guess the intention is dropping the special case? The following patch passes regtests:

That's where the mappings normally end. Example:

$ cat /proc/self/maps
2aa00000000-2aa00002000 r--p 00000000 5e:01 668061                       /usr/bin/cat
2aa00002000-2aa00006000 r-xp 00002000 5e:01 668061                       /usr/bin/cat
2aa00006000-2aa00008000 r--p 00006000 5e:01 668061                       /usr/bin/cat
2aa00008000-2aa00009000 r--p 00007000 5e:01 668061                       /usr/bin/cat
2aa00009000-2aa0000a000 rw-p 00008000 5e:01 668061                       /usr/bin/cat
2aa0000a000-2aa0002b000 rw-p 00000000 00:00 0                            [heap]
3fff7500000-3fff7557000 r--p 00000000 5e:01 926342                       /usr/lib/locale/C.utf8/LC_CTYPE
3fff7580000-3fff7581000 r--p 00000000 5e:01 658596                       /usr/lib/locale/en_US.utf8/LC_NUMERIC
3fff7600000-3fff7878000 r--p 00000000 5e:01 658592                       /usr/lib/locale/en_US.utf8/LC_COLLATE
3fff7880000-3fff7881000 r--p 00000000 5e:01 786499                       /usr/lib/locale/en_US.utf8/LC_TIME
3fff7900000-3fff7901000 r--p 00000000 5e:01 786497                       /usr/lib/locale/en_US.utf8/LC_MONETARY
3fff7980000-3fff7981000 r--p 00000000 5e:01 658594                       /usr/lib/locale/en_US.utf8/LC_MESSAGES/SYS_LC_MESSAGES
3fff7a00000-3fff7a01000 r--p 00000000 5e:01 658611                       /usr/lib/locale/en_US.utf8/LC_PAPER
3fff7a80000-3fff7a81000 r--p 00000000 5e:01 658595                       /usr/lib/locale/en_US.utf8/LC_NAME
3fff7b00000-3fff7b01000 r--p 00000000 5e:01 786493                       /usr/lib/locale/en_US.utf8/LC_ADDRESS
3fff7b80000-3fff7b81000 r--p 00000000 5e:01 786498                       /usr/lib/locale/en_US.utf8/LC_TELEPHONE
3fff7c00000-3fff7c34000 r--p 00000000 5e:01 655337                       /usr/lib64/libc.so.6
3fff7c34000-3fff7d6e000 r-xp 00034000 5e:01 655337                       /usr/lib64/libc.so.6
3fff7d6e000-3fff7dc5000 r--p 0016e000 5e:01 655337                       /usr/lib64/libc.so.6
3fff7dc5000-3fff7dc6000 ---p 001c5000 5e:01 655337                       /usr/lib64/libc.so.6
3fff7dc6000-3fff7dca000 r--p 001c5000 5e:01 655337                       /usr/lib64/libc.so.6
3fff7dca000-3fff7dcc000 rw-p 001c9000 5e:01 655337                       /usr/lib64/libc.so.6
3fff7dcc000-3fff7dd4000 rw-p 00000000 00:00 0 
3fff7e00000-3fff7e01000 r--p 00000000 5e:01 786496                       /usr/lib/locale/en_US.utf8/LC_MEASUREMENT
3fff7e80000-3fff7e87000 r--s 00000000 5e:01 654146                       /usr/lib64/gconv/gconv-modules.cache
3fff7f00000-3fff7f01000 r--p 00000000 5e:01 786495                       /usr/lib/locale/en_US.utf8/LC_IDENTIFICATION
3fff7f80000-3fff7f82000 r--p 00000000 5e:01 655334                       /usr/lib/ld64.so.1
3fff7f82000-3fff7fa2000 r-xp 00002000 5e:01 655334                       /usr/lib/ld64.so.1
3fff7fa2000-3fff7fad000 r--p 00022000 5e:01 655334                       /usr/lib/ld64.so.1
3fff7fad000-3fff7faf000 r--p 0002c000 5e:01 655334                       /usr/lib/ld64.so.1
3fff7faf000-3fff7fb1000 rw-p 0002e000 5e:01 655334                       /usr/lib/ld64.so.1
3fff7fd3000-3fff7ffb000 rw-p 00000000 00:00 0 
3fffffda000-3ffffffb000 rw-p 00000000 00:00 0                            [stack]
3ffffffc000-3ffffffe000 r--p 00000000 00:00 0                            [vvar]
3ffffffe000-40000000000 r-xp 00000000 00:00 0                            [vdso]

A higher address should work as well. I will test 0x50000000000ULL and 0x60000000000ULL and let you know the result.

Oh, right. I also found def U6Imm.

• Delete other U6Imm-related definitions.

https://reviews.llvm.org/D147521

This now causes: https://lab.llvm.org/buildbot/#/builders/57/builds/25796

• Drop imm32zx6.

I can't really comment on the implementation, but the examples in the testcase look good to me.

• Exclude Android and BSD.

Ping.

• Compute datasz only when necessary.

• Do not touch memory if the version is incorrect.
• Test this.

Ping.

Ping. Is there anything else that needs improvement?

• Do not define the constants, the kernel provides them since v2.6.
• Rewrite the test in C.

• Added a test with real args/argv.

• Drop strict_string_checks.
• Extend the testcase to check a more realistic use case.

I don't think so. First of all, this is on another architecture; and then I tried running this test on both x86_64 and s390x with your change, and it succeeded in both cases for me.

Thanks for catching this! LGTM.

lldb-arm-ubuntu (https://lab.llvm.org/buildbot/#/builders/17/builds/9015) fails with:

llvm-clang-x86_64-expensive-checks-debian (https://lab.llvm.org/buildbot/#/builders/16/builds/13741) fails with: