PPCPorts: the only ports you ever need on a PowerPC

[barracuda156]

Thanks for directing me to the correct commit. The one I had found had some patches that didn't apply, so I didn't lose much time trying. I will wait to try this until late Tuesday and check on how it went late Friday - it sound like it may take that long to build on a Powerbook running in reduced mode.

A few updates: I switched to rustc 1.39.0, since 1.29.x has a stupid bug with creating directories, and I do not know how to fix it (and upstream commit claiming to fix it did not). 1.39.0 requires LLVM-9, however incompatibilities with LLVM-7 are minimal and trivial, and should be fixed in this patch.
This is a more recent version of portfile – however I did not commit the latest changes which I had locally, and won’t be able to until end of month.
The current status is that everything builds up to the end of bootstrap stage (so rustc is built), however something does not work with creating LLVM target, so it is not usable at the moment.

Possibly, target specs has to be modified, but I am not yet sure. It takes a lot of time to rebuild (and it does not allow to recompile a single file).

mrustc itself appears functional (since it compiles rust in bootstrap stage), but I have no idea what it can build aside of rustc bootstrap.

In other news, I got current Poppler to build configured with -DENABLE_GPGME=OFF \ and without the boost variant (-boost). I know poppler is not in PowerPC ports, so I won't make a pull request, but I wanted to note this somewhere in case someone else on Tiger needs modern poppler. The fallback that builds with the default variants and configure options was too old to build evince.

We could probably just add poppler, if changes for 10.4 are minimal and the same version is used. Alternatively, poppler-legacy can be added, and then used as a drop-in.

However, if you expect many of such instances, it is perhaps better to have a Tiger-specific tree (either with additional ports or as an extra overlay, only with 10.4-specific ones). What do you think? It will probably be easier for you (or whomever maintains ports for 10.4), and it won’t add stuff for later systems unnecessarily.

Speaking of ports not in PowerPC ports, I also made a patch that allows current libxkbcommon-x11 from Macports to build, which hopefully will reduce broken files during rebuilds. The patch is attached.

Please submit the PR with it. But let’s consider options for how to support 10.4 better. (I am not politely implying that a separate tree is the only choice, it is discussable.)

 

[Forest Expertise]

A few updates: I switched to rustc 1.39.0, since 1.29.x has a stupid bug with creating directories, and I do not know how to fix it (and upstream commit claiming to fix it did not). 1.39.0 requires LLVM-9, however incompatibilities with LLVM-7 are minimal and trivial, and should be fixed in this patch.
This is a more recent version of portfile – however I did not commit the latest changes which I had locally, and won’t be able to until end of month.
The current status is that everything builds up to the end of bootstrap stage (so rustc is built), however something does not work with creating LLVM target, so it is not usable at the moment.

View attachment 2580246

Possibly, target specs has to be modified, but I am not yet sure. It takes a lot of time to rebuild (and it does not allow to recompile a single file).

mrustc itself appears functional (since it compiles rust in bootstrap stage), but I have no idea what it can build aside of rustc bootstrap.



We could probably just add poppler, if changes for 10.4 are minimal and the same version is used. Alternatively, poppler-legacy can be added, and then used as a drop-in.

However, if you expect many of such instances, it is perhaps better to have a Tiger-specific tree (either with additional ports or as an extra overlay, only with 10.4-specific ones). What do you think? It will probably be easier for you (or whomever maintains ports for 10.4), and it won’t add stuff for later systems unnecessarily.



Please submit the PR with it. But let’s consider options for how to support 10.4 better. (I am not politely implying that a separate tree is the only choice, it is discussable.)

Thank you for your work on mrustc again - it looks like I should prioritize checking that llvm 7.1 for powerpc works on Tiger before trying to catch up with you on mrustc.
By my count, there are eight important ports that need fixes for Tiger and are not currently present in PowerPC ports: poppler, libxkbcommon-x11, gnupg2, bash, openldap, gpgme, AtomicParsley, and orc. Poppler is fortunately extremely minimal (one configure option is added). I will make a PR for it at some point - I need to fix the other PRs I have opened first lol. I opened an issue with my perspective on Tiger's biggest current problems here: https://github.com/macos-powerpc/powerpc-ports/issues/25
If eight extra ports for Tiger is too many, AtomicParsley and orc could be postponed - that would leave six ports specifically for Tiger. How many ports added just for Tiger fixes are too many? I will try to prioritize once you let me know.
I considered a separate tree, but I think it would be nice for users if everything was in the same place. My understanding is also that having multiple contributors and more pull request activity increases project visibility - and I do think PowerPC ports is a great project that should be more visible. I think having a separate tree for Tiger decreases visibility for both. Ultimately, it is your call as the creator of PowerPC ports, and I will respect what you think is best.

 

[barracuda156]

Thank you for your work on mrustc again - it looks like I should prioritize checking that llvm 7.1 for powerpc works on Tiger before trying to catch up with you on mrustc.

If your aim is rustc, then LLVM is a must at the moment. (gcc codegen is not yet ready, AFAIK, and not implemented via mrustc at all.)
mrustc on its own does not use LLVM.

If eight extra ports for Tiger is too many, AtomicParsley and orc could be postponed - that would leave six ports specifically for Tiger. How many ports added just for Tiger fixes are too many? I will try to prioritize once you let me know.
I considered a separate tree, but I think it would be nice for users if everything was in the same place. My understanding is also that having multiple contributors and more pull request activity increases project visibility - and I do think PowerPC ports is a great project that should be more visible. I think having a separate tree for Tiger decreases visibility for both. Ultimately, it is your call as the creator of PowerPC ports, and I will respect what you think is best.

Well, we can add these first and see how it goes. Moving a subset of ports to a branch is always doable, if needed.
When adding a new port, please start with a commit that copies existing MacPorts’ port with no changes and then modify it in a following commit.