0

I'm trying to install qml package in kdb, and encounters the compiling error selecting C floating-point options... not found -ffloat-store is assumed by some libraries.

I've found this post three years ago asking the same question. compiling qml in macOS kdb+/q (CFLAG?) Yet the solution provided doesn't work for me. After brew install gcc and ./configure CC=gcc I still got the same error message.

Besides that, I also have another issue, the configure always recognizes kdb as m64 and version 3. However What I have is m32 and version 4. Does anyone know how I can manually change it? thanks.

Lynette
  • 13
  • 4
  • I found out the version number of gcc by `brew info gcc`, then specify it as `./configure CC=gcc-10` The configure went through. The make still has error, I presume it's caused by the configure set kdb as m64 and version 3. Is there any way to change it? I still haven't figured out. Thanks. – Lynette Aug 15 '20 at 13:17
  • have you tried modifying the compile script to -m32 instead of -m64 like the comment on that question suggests? – Matt Moore Aug 15 '20 at 16:26
  • I've run `./configure`, then opened the intermediate file config.mk, change two variables KXVER and FLAGS from m64 to m32. Then `make` went through, `make test` didn't go through,`make install` also went through. It looks it's installed successfully. I tried to load it. `\l qml.q`, then I got error messages `'qml: loading qml_sin(): dlsym(0x7a70ff30, qml_sin): symbol not found`. Not sure what can I look into next. – Lynette Aug 15 '20 at 17:14
  • I finally got qml running on my Mac. The trick is rebuilding Blas and Lapack, then it can wrap qml, Blas and Lapack as a 32bit library, then 32bit kdb+ can load it. I'm hopeful this can be useful for other kdb+developers. Also, gcc-10 is different from previous versions, the default is -fno-common, need to change it back to -fcommon. https://gcc.gnu.org/gcc-10/changes.html#c Thank you very much for those trying to help! – Lynette Aug 19 '20 at 17:18

0 Answers0