A tracking issue? Roadmap? TODO list? Notes? of sorts for things that need to be handled in order to get Nixpkgs¹ working natively on powerpc64-linux (the big-endian variant, not powerpc64le-linux). I could immediately bulk-submit PRs for all of these (and some more issues I already debugged a year ago during a previous effort), but submitting fixes for a not-yet-functioning stdenv feels abit silly.

The following things need to be addressed for stdenv to build & work. I would consider this issue "done" and closable once these are done, as anything after stdenv can just be fixed incrementally. No need to keep this ticket open for all eternity. 🙂

• The dynamic linker for powerpc64-linux is broken, preventing cross-compiled freshBootstrapTools.bootstrapFiles from being usable.
  Waiting for bintools-wrapper: fix dynamic linker for powerpc64 big-endian #247682 to fix this.
• The Linux stdenv needs bootstrapFiles to be added for powerpc64-linux so native bootstrapping can begin.
  Previous attempt: pkgs/stdenv/linux: add powerpc64 bootstrap files #177001
  Current attempt: pkgs/stdenv/linux: Add powerpc64 bootstrap-files #255451
• openssl needs a configureScript entry to be added for powerpc64-linux, otherwise it warns about & attempts to do a 32-bit build.
  Required configureScript: "./Configure linux-ppc64"
  openssl: Add configureScript entry for powerpc64-linux #295268
• pcre needs a patch from Void Linux for its tests to pass for me: void-linux/void-packages:srcpkgs/pcre/patches/ppc-icache-flush.patch
  From the patch comments, this one might be due to the age of my CPU (POWER4)? Haven't checked if pcre2 has the same problem.
  pcre: Fetch patch for powerpc64-linux #295442

These are some additional problems I've encountered after reaching stdenv and attempting to build my way up so some graphical applications. I'll document things here as I encounter them while the above list isn't done, just so I don't forget about them. Once stdenv works on master, I'll just submit PRs for these & future problems.

• llvmPackages*.llvm: All MCJIT and OrcMCJIT tests fail.
  (I haven't actually tested any llvm other than the default one due to long compile times, but with a failure this massive I don't see how this could be version-specific.)
  Currently just disabling doCheck, not sure about the why's & how's on this.
• openexr & openexr_3: Hitting all the same test failures as Fedora: Some tests fail for s390x arch AcademySoftwareFoundation/openexr#1175, and ones reported in other upstream issues over the years.
  Will disable the tests since it doesn't look like full fixes for these are in sight, although I wonder if openexr* in general is broken on big-endian and should instead just be marked as such?
• tpm2-tss: Same unit tests failures as reported upstream on s390x (s390x test failures tpm2-software/tpm2-tss#2531), but additionally all integration tests get stuck in a retry loop and eventually time out.
  Haven't looked into what's going on with the integration tests yet, but same thoughts on this as with openexr: Disabling tests for now, but possibly more of a broken situation.

ERROR: test/integration/sys-asymmetric-encrypt-decrypt
======================================================

Trying to start simulator mssim
Starting simulator on port 2218
successfully started daemon: tpm_server with PID: 95717
/build/source
simulator PID: 95717
Port conflict? Cleaning up PID: 95717
./script/int-log-compiler-common.sh: line 246: kill: (95717) - No such process
Failed to start simulator: port 2218 or  2219 probably in use. Retrying in 2.
Starting simulator on port 15806
successfully started daemon: tpm_server with PID: 95767
/build/source
simulator PID: 95767
Port conflict? Cleaning up PID: 95767
./script/int-log-compiler-common.sh: line 246: kill: (95767) - No such process
Failed to start simulator: port 15806 or  15807 probably in use. Retrying in 4.
Starting simulator on port 30400
successfully started daemon: tpm_server with PID: 95799
/build/source
simulator PID: 95799
Port conflict? Cleaning up PID: 95799
./script/int-log-compiler-common.sh: line 246: kill: (95799) - No such process
Failed to start simulator: port 30400 or  30401 probably in use. Retrying in 8.
Starting simulator on port 31390
successfully started daemon: tpm_server with PID: 95831
/build/source
simulator PID: 95831
Port conflict? Cleaning up PID: 95831
./script/int-log-compiler-common.sh: line 246: kill: (95831) - No such process
Failed to start simulator: port 31390 or  31391 probably in use. Retrying in 16.
Starting simulator on port 6842
successfully started daemon: tpm_server with PID: 95863
/build/source
simulator PID: 95863
Port conflict? Cleaning up PID: 95863
./script/int-log-compiler-common.sh: line 246: kill: (95863) - No such process
Failed to start simulator: port 6842 or  6843 probably in use. Retrying in 32.
Starting simulator on port 59214
successfully started daemon: tpm_server with PID: 95895
/build/source
simulator PID: 95895
Port conflict? Cleaning up PID: 95895
./script/int-log-compiler-common.sh: line 246: kill: (95895) - No such process
Failed to start simulator: port 59214 or  59215 probably in use. Retrying in 64.
TPM20TEST_TCTI_NAME=socket
TPM20TEST_DEVICE_FILE=
TPM20TEST_SOCKET_ADDRESS=127.0.0.1
TPM20TEST_SOCKET_PORT=59214
TPM20TEST_TCTI=mssim:host=127.0.0.1,port=59214
WARNING:tcti:src/util/io.c:262:socket_connect() Failed to connect to host 127.0.0.1, port 59214: errno 111: Connection refused 
Failed to initialize tcti context: 0xa000a
TPM_StartUp failed
./script/int-log-compiler-common.sh: line 195: kill: (95895) - No such process
failed to kill daemon process with PID: 95895
ERROR test/integration/sys-asymmetric-encrypt-decrypt.int (exit status: 99)

¹: My dream would be NixOS on powerpc64-linux via a USB installer image, but that's definitely a separate thing to tackle (and something I've never dabbled into before).