Arch Linux User Repository

Issue with compiling 2021 R2 using gcc/11.2

gfortran -O1 -m64 -fdefault-integer-8 -std=legacy -lpthread ddi/test/ddi_test.f ddi/libddi.a -o ddi_test.x

ld: error: undefined symbol: _gfortran_os_error_at
>>> referenced by ddi_test.f
>>>               /tmp//cc7iR4y2.o:(gdlb_test_)
>>> referenced by ddi_test.f
>>>               /tmp//cc7iR4y2.o:(gdlb_test_)
collect2: error: ld returned 1 exit status

I'm not sure how to correct this

I can't install gamess: it lists a lot of undefined references like 'dsygst_', 'dsyevd_', 'dgetri_' and 'dsyev_'. What can I do?

I have a question regarding the OpenMP option that this package uses in the default python configuration. Is there a depend on openmp?

@xorinet According to https://stackoverflow.com/a/42326612 , I worked around that by installing lapack and lapacke and then modifying the PKGBUILD as follows:

diff --git a/PKGBUILD b/PKGBUILD
index de2ef06..2f6cdab 100644
--- a/PKGBUILD
+++ b/PKGBUILD
@@ -33,6 +33,8 @@ prepare() {

   # Blas-agnostic
   sed -i 's/-lopenblas/-lblas/g' lked
+
+  sed -i 's/$LIBXC_FLAGS/$LIBXC_FLAGS -llapacke -llapack/g' lked
 }

 build() {

(i.e. I added sed -i 's/$LIBXC_FLAGS/$LIBXC_FLAGS -llapacke -llapack/g' lked in the prepare function.)

@budgerigar Would you please fix this? (And thanks for your work all along!)

There seems to be an error while linking (2021-R1):

/usr/bin/ld: dftbx.o: in function 'diaglap_':
dftbx.f:(.text+0x7f79): undefined reference to 'dsygst_'
/usr/bin/ld: dftbx.f:(.text+0x7fbf): undefined reference to 'dsyevd_'
/usr/bin/ld: pcmcv2.o: in function 'cpcmmi_':
pcmcv2.f:(.text+0xf60d): undefined reference to 'dgetri_'
/usr/bin/ld: pcmcv2.f:(.text+0xf674): undefined reference to 'dgetri_'
/usr/bin/ld: eigen.o: in function 'gldiag_':
eigen.f:(.text+0x6da2): undefined reference to 'dsyev_'
collect2: error: ld returned 1 exit status

Any fixes? Thanks!

There is a fix for this. It's been determined that tcsh-6.22.03 is the cause of the problem. Starting at line 120 in the comp script, the if; then statement messes up the path to the source file to be compiled. Apparently there is a change for tcsh-6.22.04 that should correct this

https://bbs.archlinux.org/viewtopic.php?pid=1975493#p1975493

I'll add that I tried to build 2020R2 on Ubuntu 21.04 with:

tcsh 6.21, gfortran 10.3, make 4.3

It worked without any issues on Ubuntu 21.04. Doing the same under a fresh Arch install results in the sed error if I use this PKGBUILD. Without this PKGBUILD, the sed error is not present but the rest of the error starting from gfortran: error: libxc_empty: No such file or directory is still occurring. If I use the old ./compall method, the gfortran error similarly occurs for all modules.

I have also tried doing this in a clean chroot as well and the error persists

Thank you for looking into it

Guys, are there any ideas why sed stopped working that suddenly?

I downloaded the gamess-current.tar.gz from the website and they state the latest version as 2020R2 from September 30. I recently attempted to build this and I get a sed related error during make modules -j1

The exact error is:

sed: can't read libxc_empty: No such file or directory gfortran: error: libxc_empty: No such file or directory gfortran: fatal error: no input files compilation terminated unset echo Division by 0 make: *** [/home/User/Build/gamess/src/gamess/Makefile.in:99: /home/User/Build/gamess/src/gamess/object/libxc_empty.o] Error 1

Has something changed about the build tools since the last comment on 2020-11-15?

make 4.3 gcc-fortran 10.2 cmake 3.20.2 openblas 0.3.14-1 tcsh 6.22.03 python-jinja 2.11.3-2 python 3.9.4-1 inetutils 1.9.4-9

@asakurahao, this was done. Please rebuild the package.

Dear, I need the blas package for another software and Gamess try to change it for openblas. Is there a way to install the openblas without lose the blas?

@Gemini

OK, I see.

@SamLukeYes

They force us to get it at their website, to accept some kind of agreement, then send to you via email after evaluation. I believe this is explained somewhere in PKGBUILD. Once you get it, place it in the same directory before running pacman.

The PKGBUILD doesn't provide the URL of gamess-current.tar.gz. Where should I fetch it?

@berquist You did! Thanks! :)

Now if only I had an interface to gamess, since they're also failing to install and their packagers/maintainers are not as efficient as you guys. Hopefully I'll get a reply to my comments soon. Again, thanks!

@Gemini I believe I've fixed it. It should now also build properly under a chroot.

@a.kudelin nope. I could paste here, if you like.

I got the gamess source yesterday, could they have updated something that broke it? Also, on my previous post, that "bash: Compilation: command not found" is suspicious, it seems to lack that actvte.x tool, and asking to run config there, in order to get it.

@Gemini, I cannot reproduce the problem. Did you alter PKGBUILD somehow?

Got the source ok and the DDI compilation worked. The following compilation didn't start:

DDI compilation ended successfully.
Fri Apr  3 21:50:06 UTC 2020
16.290u 4.193s 1:27.42 23.4%    0+0k 16518+19784io 76pf+0w
Compilation of /home/user/builds/gamess/src/gamess/tools/actvte.x requires running /home/user/builds/gamess/src/gamess/config
bash: Compilation: command not found
make: *** [/home/user/builds/gamess/src/gamess/Makefile.in:95: /home/user/builds/gamess/src/gamess/tools/actvte.x] Error 127
==> ERROR: A failure occurred in build().

@berquist a.kudelin thanks for your help.

@budgerigar Sorry I didn't see this earlier (2 years ago). I have a PKGBUILD for 2018R3 that works so would be fine to help maintain.

@budgerigar Could you add me as a maintainer?