rPackages.Rhisat2: fix build

[Kupac]

Description of changes

Add required commands

Things done

• Built on platform(s)
  • x86_64-linux
  • aarch64-linux
  • x86_64-darwin
  • aarch64-darwin
• For non-Linux: Is sandboxing enabled in nix.conf? (See Nix manual)
  • sandbox = relaxed
  • sandbox = true
• Tested, as applicable:
  • NixOS test(s) (look inside nixos/tests)
  • and/or package tests
  • or, for functions and "core" functionality, tests in lib/tests or pkgs/test
  • made sure NixOS tests are linked to the relevant packages
• Tested compilation of all packages that depend on this change using nix-shell -p nixpkgs-review --run "nixpkgs-review rev HEAD". Note: all changes have to be committed, also see nixpkgs-review usage
• Tested basic functionality of all binary files (usually in ./result/bin/)
• 24.05 Release Notes (or backporting 23.05 and 23.11 Release notes)
  • (Package updates) Added a release notes entry if the change is major or breaking
  • (Module updates) Added a release notes entry if the change is significant
  • (Module addition) Added a release notes entry if adding a new NixOS module
• Fits CONTRIBUTING.md.

---

Add a 👍 reaction to pull requests you find important.

[b-rodrigues]

Doesn't build on aarch64-linux:

g++: error: unrecognized command-line option '-m64'
g++: error: unrecognized command-line option '-msse2'
make: *** [Makefile:363: hisat2-build-s] Error 1
ERROR: compilation failed for package 'Rhisat2'
* removing '/nix/store/c4p14lc9a2qgj8pzr2xdcnaf44dgmb1c-r-Rhisat2-1.18.0/library/Rhisat2'
error: builder for '/nix/store/vchryn0rxw0jciqq2wh0rs4xf8br96py-r-Rhisat2-1.18.0.drv' failed with exit code 1;
       last 10 log lines:
       > -fno-strict-aliasing -DHISAT2_VERSION="\"2.2.1\"" -DBUILD_HOST="\"localhost\"" -DBUILD_TIME="\"Tue Apr  9 19:51:52 UTC 2024\"" -DCOMPILER_VERSION="\"`/nix/store/49nb1m1ysq0vx31qp94alpmi6ccsq9l3-gcc-wrapper-13.2.0/bin/g++ -v 2>&1 | tail -1`\"" -D_LARGEFILE_SOURCE -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE  -DBOWTIE_MM  -DUSESIMDE -DBOWTIE2 -DNDEBUG -Wall -DMASSIVE_DATA_RLCSA \
       > -I. -I third_party \
       > -o hisat2-build-s hisat2_build.cpp \
       > ccnt_lut.cpp ref_read.cpp alphabet.cpp shmem.cpp edit.cpp gfm.cpp reference.cpp ds.cpp multikey_qsort.cpp limit.cpp random_source.cpp tinythread.cpp diff_sample.cpp hisat2_build_main.cpp \
       > -lpthread
       > g++: error: unrecognized command-line option '-m64'
       > g++: error: unrecognized command-line option '-msse2'
       > make: *** [Makefile:363: hisat2-build-s] Error 1
       > ERROR: compilation failed for package 'Rhisat2'
       > * removing '/nix/store/c4p14lc9a2qgj8pzr2xdcnaf44dgmb1c-r-Rhisat2-1.18.0/library/Rhisat2'
       For full logs, run 'nix log /nix/store/vchryn0rxw0jciqq2wh0rs4xf8br96py-r-Rhisat2-1.18.0.drv'.

guess this would need some specific fix for aarch64-linux, but this could be addressed in another PR.

[Kupac]

Nah, we should try to fix it here. Probably those command line options dont make sense on aarch.

[Kupac]

But TBH I don't know how to fix these compiler issues

[b-rodrigues]

I'll give it a shot

[jbedo]

Makes sense, -m64 and -msse2 are x86 specific. Upstream now support aarch64, you could backport the patch: fmicompbio/Rhisat2@a0f27b0

[b-rodrigues]

So I was able to get it to work on my Rpi 5 by adding this:

    Rhisat2 = old.Rhisat2.overrideAttrs (attrs: {
      postPatch = ''
        substituteInPlace "src/Makefile" \
          --replace-fail "BITS_FLAG = -m64" "BITS_FLAG =" \
          --replace-fail "SSE_FLAG=-msse2" "SSE_FLAG="
      '';
      env = (attrs.env or { }) // {
        NIX_CFLAGS_COMPILE = attrs.env.NIX_CFLAGS_COMPILE + " -w";
      };
    });

but I don't know:

• how to just use this on aarch64-linux
• if this would be the most elegant way of doing it

the NIX_CFLAGS_COMPILE overwrite is because of this warning:

ERROR: compilation failed for package 'Rhisat2'
* removing '/nix/store/9gbixqccpb29hs9h4wchm8007gy917p4-r-Rhisat2-1.18.0/library/Rhisat2'
error: builder for '/nix/store/jpamm9m9dzlavzyby6iswc71vd2h6j3i-r-Rhisat2-1.18.0.drv' failed with exit code 1;
       last 10 log lines:
       > reference.cpp: In member function 'int BitPairReference::getStretch(uint32_t*, size_t, size_t, size_t) const':
       > reference.cpp:522:26: warning: variable 'origBufOff' set but not used [8;;https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#index-Wunused-but-set-variable-Wunused-but-set-variable8;;]
       >   522 |                 uint64_t origBufOff = bufOff;
       >       |                          ^~~~~~~~~~
       > reference.cpp:516:14: warning: variable 'binarySearched' set but not used [8;;https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#index-Wunused-but-set-variable-Wunused-but-set-variable8;;]
       >   516 |         bool binarySearched = false;
       >       |              ^~~~~~~~~~~~~~
       > make: *** [Makefile:363: hisat2-build-s] Error 1
       > ERROR: compilation failed for package 'Rhisat2'
       > * removing '/nix/store/9gbixqccpb29hs9h4wchm8007gy917p4-r-Rhisat2-1.18.0/library/Rhisat2'
       For full logs, run 'nix log /nix/store/jpamm9m9dzlavzyby6iswc71vd2h6j3i-r-Rhisat2-1.18.0.drv'.

so I followed the advice linked.

> Rhisat2::hisat2_version()
[1] "/nix/store/130gkb8vcj54g4jyp7v00xl1w4qq1cmr-r-Rhisat2-1.18.0/library/Rhisat2/hisat2 version 2.2.1"
[2] "64-bit"                                                                                           
[3] "Built on localhost"                                                                               
[4] "Fri Apr 12 10:22:13 UTC 2024"                                                                     
[5] "Compiler: gcc version 13.2.0 (GCC) "                                                              
[6] "Options: -O3   -funroll-loops -g3 -std=c++11"                                                     
[7] "Sizeof {int, long, long long, void*, size_t, off_t}: {4, 8, 8, 8, 8, 8}"

[b-rodrigues]

I see that there’s also some macOS-specific stuff in the Makefile, for curiosity’s sake, @philipp-baumann could you try to build this on darwin? Might as well fix it completely in one go 😃

[jbedo]

I think it'd be better to backport the patch, you're essentially recreating it in your nix expression.

[b-rodrigues]

I think it'd be better to backport the patch, you're essentially recreating it in your nix expression.

ok, here it is:

b-rodrigues@3c4e1f8

I don't know how to add this to this PR, I think the easiest is for @Kupac to simply add it?

[Kupac]

@ofborg build rPackages.Rhisat2

[Kupac]

OK, now darwin builds have various errors. aarch64-darwin complains about msse2, but didn't you remove that flag on aarch already?

[b-rodrigues]

That's what the patch does but there's two other files that I didn't patch which were changed in the original commit, but it didn't seem to be related

fmicompbio/Rhisat2@a0f27b0

[jbedo]

Source the patch from github.

[Kupac]

@ofborg build rPackages.Rhisat2

[Kupac]

@jbedo I can't resolve the changes requested once again. I maybe missing something on the github gui.

[philipp-baumann]

on aarch64-darwin i get

       > /nix/store/0m04cmy32vs1686pzjqajsh50arbv8m4-clang-wrapper-16.0.6/bin/cc -I"/nix/store/6s73b2138xz0k5xc2w9bf8c5ic74js46-R-4.3.2/lib/R/include" -DNDEBUG -DLIBXML -I/nix/store/05g7i28lm20a7wp3hn5pgmkxwqlxqxzs-libxml2-2.12.5-dev/include/libxml2 -DUSE_EXTERNAL_SUBSET=1 -DROOT_HAS_DTD_NODE=1 -DNO_CHECKED_ENTITY_FIELD=1 -DDUMP_WITH_ENCODING=1 -DUSE_XML_VERSION_H=1 -DXML_ELEMENT_ETYPE=1 -DXML_ATTRIBUTE_ATYPE=1 -DNO_XML_HASH_SCANNER_RETURN=1 -DLIBXML_NAMESPACE_HAS_CONTEXT=1 -DHAVE_R_CETYPE_T=1 -D__XMLSEC_FUNCTION__=__func__ -DXMLSEC_NO_SIZE_T -DXMLSEC_NO_GOST=1 -DXMLSEC_NO_GOST2012=1 -DXMLSEC_DL_LIBLTDL=1 -I/nix/store/mkfw27hp2c5fybgqkgvfh4mib9wp8y1n-xmlsec-1.2.34-dev/include/xmlsec1 -I/nix/store/05g7i28lm20a7wp3hn5pgmkxwqlxqxzs-libxml2-2.12.5-dev/include/libxml2 -I/nix/store/189j4cpdxvrcr0s2841yf2vpirw9l11h-libxslt-1.1.39-dev/include -DXMLSEC_CRYPTO_DYNAMIC_LOADING=1 -DHAVE_XML_WITH_ZLIB=1 -DHAVE_XML_HAS_FEATURE=1 -DUSE_R=1 -D_R_=1  -DHAVE_VALIDITY=1 -DXML_REF_COUNT_NODES=1 -DHAVE_LIBXML_SEC=1 -I. -DLIBXML2=1  -isystem /nix/store/rjgv94p1nnj02v0f9r2jpann9x02b4vw-libcxx-16.0.6-dev/include/c++/v1    -fPIC  -g -O2  -c XMLEventParse.c -o XMLEventParse.o
       > XMLEventParse.c:618:31: error: incompatible function pointer types assigning to 'xmlStructuredErrorFunc' (aka 'void (*)(void *, const struct _xmlError *)') from 'void (void *, xmlErrorPtr)' (aka 'void (void *, struct _xmlError *)') [-Wincompatible-function-pointer-types]
       >      xmlParserHandler->serror = RS_XML(structuredErrorHandler);
       >                               ^ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
       > 1 error generated.
       > make: *** [/nix/store/6s73b2138xz0k5xc2w9bf8c5ic74js46-R-4.3.2/lib/R/etc/Makeconf:191: XMLEventParse.o] Error 1
       > ERROR: compilation failed for package 'XML'
       > * removing '/nix/store/8rhiqrqp3dhshbsh517iqmar6lad6pdf-r-XML-3.99-0.16/library/XML'
       For full logs, run 'nix log /nix/store/v0f32bk3np8day46ghz3fmyi7xs818p6-r-XML-3.99-0.16.drv'.
error: 1 dependencies of derivation '/nix/store/j5893qi32x0wgbkc3nhwpfwz6xihwz7h-r-biomaRt-2.58.0.drv' failed to build
error: 1 dependencies of derivation '/nix/store/6nbc7i0vn6lpik375jb67s0cnd5ddxz1-r-rtracklayer-1.62.0.drv' failed to build
error (ignored): error: cannot unlink '/private/tmp/nix-build-r-igraph-1.6.0.drv-0/igraph': Directory not empty
error: 1 dependencies of derivation '/nix/store/ikrmwgy898lspgp47nz02v0568vam664-r-GenomicFeatures-1.54.1.drv' failed to build
error: 1 dependencies of derivation '/nix/store/sm6lkknv6l0lsadik0syg8x3d4ypqnz3-r-Rhisat2-1.18.0.drv' failed to build

nixpkgs on  fix_Rhisat2 [$?] took 20.2s 

[b-rodrigues]

on aarch64-darwin i get

looks like rPackages.XML needs fixing

[philipp-baumann]

on aarch64-darwin i get

looks like rPackages.XML needs fixing

yeah that one would have quite a bit of impact for macOS builds, still a lot reverse depends to be expected. https://stat.ethz.ch/pipermail/r-package-devel/2024q1/010374.html

[Kupac]

Weird, it's not the same error message as ofborg's. It clearly says * DONE (XML). Are you sure it's related?

[philipp-baumann]

Weird, it's not the same error message as ofborg's. It clearly says * DONE (XML). Are you sure it's related?

rPackages.XML alone does not build on my M2 machine for the code base at this PR stage. Maybe ofborg does not rebuild XML package (maybe updated?)

[Kupac]

I think ofborg should work from the same commit as you, no? In theory, at least. For me, so far, the evaluation log was always identical to ofborg's.