<timokau[m]>
Do we have a standard way to refer to the "link name" of a library?
<Mic92>
We are getting many darwin changes lately. Did nixpkgs become more popular there?
<timokau[m]>
Here the issue is that we want to abstract over the blas lib used, which needs to be specified with its linker flag `-lopenblas` or `-lblas`
<timokau[m]>
Mic92: not sure, in this case the user is trying to make sage build and making a lot of PRs fixing up dependencies because of that
<Mic92>
timokau[m]: this was more a general finding.
<timokau[m]>
The `linkName` thing he does seems nice and may be worth integrating directly into the blas libs if we don't already have a standard for that
<timokau[m]>
Mic92: certainly makes sense for me that nix would be popular for mac users
<Mic92>
timokau[m]: you can also add a passthrough variable in the blas library itself.
<Mic92>
if there different consumsers for this
<Mic92>
I don't think we have this problem that often.
<timokau[m]>
Mic92: yeah that is what I meant by "integrating it directly into the blas libs"
<timokau[m]>
Just checking if there already is an existing standard before suggesting that
<timokau[m]>
I think for standard that have different implementations like blas it makes sense
<Mic92>
passthru instead of passthrough its called
<Mic92>
I would not over-engineer it unless, the problem occurs in many places.
<timokau[m]>
Mic92: well `-l*blas` occurs in 33 places, all of which could probably make use of a configurable blas implementation
<timokau[m]>
In practice its probably not necessary to make that configurable in most cases though, so I don't know if the generic mechanism is better
<Mic92>
we have such an abstraction called unixtools
<timokau[m]>
What is that?
<Mic92>
also I would not put blas there
<Mic92>
It is the same thing for common commandline tools
<Mic92>
things like utillinux
<timokau[m]>
Mic92: that sounds like a different thing
<timokau[m]>
openblas usually is platform agnostic and should work on darwin
<timokau[m]>
Its just that it triggers some weird error in this case
init_6 has quit [Ping timeout: 272 seconds]
<layus>
I wonder how to test modifications to cc-wrapper without rebuilding all the world ?
<ekleog>
infinisil: stupid question: does {^_^} keep things-to-tell across downups? I had launched one to Sonarpulse that hasn't triggered yet, so that'd be to know whether I should resend it :)
<infinisil>
ekleog: something with channel modes gchristensen set yesterday apparently made {^_^} exit, I can't do anything about this myself
<infinisil>
But yeah it keeps these ,tell's between restarts
<ekleog>
oh :'(
<ekleog>
and 'k, thanks!
<ekleog>
I guess {^_^} is not registered, and it was forwarded to #nixos-unregistered :) (at least, it's there)