![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
TMI: Feminist Friday
Jul. 26th, 2013 07:50 pmToday I headed to the San Francisco office for a day of co-working in the community space there, with Lukas, Liz, and some folks from the community. It was fun and I want to go again!
On the train on the way there, I finally got my extern-mod test case to compile successfully. Among other things, it turns out the Mac OS X dynamic linker isn't very happy when you rename a .dylib file. So now, rustpkg is doing what it should have been doing and preserving the long name (libwhatever-hash-version.dylib) in the installed library, instead of building build/packagename/libwhatever-hash-version.dylib to lib/libwhatever-version.dylib). I'm not sure why I was ever doing it that way, except my own ignorance!
One thing that I still need to do is make rustc print out an error message that suggests using rustpkg to download the package, if the user writes something like extern mod github.com/catamorphism/whatever and there's no locally cached copy. rustc isn't responsible for downloading anything off the network, git is, but currently it'll just treat it like any other path that doesn't exist.
I also rebased my branch to use workcache for avoiding recompilation, now that Graydon's pull request to make the workcache context sendable landed. I'm sure it's not going to work yet, since I just did the minimum necessary to invoke the workcache and make it typecheck. Still waiting for the build there.
Finally, I got my pull request for handling non-numeric tags into the queue. Graydon pointed out I shouldn't assume a given refspec is meant to be a tag, and that git can do a pretty good job of working out whether it's a tag, revision, or so on. So I fixed that, and now it's just waiting for bors to get to it.
On the train on the way there, I finally got my extern-mod test case to compile successfully. Among other things, it turns out the Mac OS X dynamic linker isn't very happy when you rename a .dylib file. So now, rustpkg is doing what it should have been doing and preserving the long name (libwhatever-hash-version.dylib) in the installed library, instead of building build/packagename/libwhatever-hash-version.dylib to lib/libwhatever-version.dylib). I'm not sure why I was ever doing it that way, except my own ignorance!
One thing that I still need to do is make rustc print out an error message that suggests using rustpkg to download the package, if the user writes something like extern mod github.com/catamorphism/whatever and there's no locally cached copy. rustc isn't responsible for downloading anything off the network, git is, but currently it'll just treat it like any other path that doesn't exist.
I also rebased my branch to use workcache for avoiding recompilation, now that Graydon's pull request to make the workcache context sendable landed. I'm sure it's not going to work yet, since I just did the minimum necessary to invoke the workcache and make it typecheck. Still waiting for the build there.
Finally, I got my pull request for handling non-numeric tags into the queue. Graydon pointed out I shouldn't assume a given refspec is meant to be a tag, and that git can do a pretty good job of working out whether it's a tag, revision, or so on. So I fixed that, and now it's just waiting for bors to get to it.