On Tue, 9 Jul 2024, matthew green wrote:
i wonder if a better default would be ultrasparct1? sun4v-cpu will always be one of those or better, won't it? is there another way to find out what actual CPU is available? the above isn't wrong, but perhaps it can be better. since this looks like a general solaris vs mpfr/gmp issue, can you raise it upstream as well? they might have a better idea again.
I just grabbed the sources from the gmplib.org site (6.2.1 which matches what we have in current netbsd and 6.3.0 which is the latest official gmp release).
Running "./configure" reveals the following: 6.2.1: configure: summary of build options: Version: GNU MP 6.2.1 Host type: sparc-sun-solaris2.11 ABI: 32 Install prefix: /usr/local Compiler: gcc Static libraries: yes Shared libraries: yes 6.3.0: configure: summary of build options: Version: GNU MP 6.3.0 Host type: ultrasparc-sun-solaris2.11 ABI: 64 Install prefix: /usr/local Compiler: gcc Static libraries: yes Shared libraries: yesSo the latest release (6.3.0) seems to be able to detect the host type and ABI properly.
I guess upgrading the netbsd codebase to use 6.3.0 is not an high priority option for this issue with building on a Solaris sun4v/ldom host.
I will have a look at the changes from 6.2.1 to 6.3.0 to see if I can determine how the host type detection logic has changed.