• And bump port revision.

Add prefix to OpenBLAS headers with common names (like config.h or common.h).

Sorry for forgetting about this. I didn't added it to my TODO list.

The current port doesn't provide libtensorflow.so, which is required for Haskell bindings, but it's better to build upon that, rather that this solution, which requires repackaging.

Thanks for your submission, and sorry for this taking so long.

In D16623#459872, @uzsolt_uzsolt.hu wrote:

In D16623#459817, @arrowd wrote:

Just out of curiosity, what's the rationale for this package split?

The "old" texlive-texmf installed size is about 1.35GiB without any documentation - it's plus 1.58GiB so they're about 3GiB and for example I don't use many-many (useful) packages (styles, classes, fonts, etc.).
I think the (new) texlive-texmf-publishers (for example) is needed only some people so it would be about 230MiB unnecessary files for many people. For my daily work I need only 10 texlive-texmf-* from all 36 (about one third). So I (and other texlive-users too) need install only the necessary packages (plus some similar) - save storage, bandwidth,... The test of texmf-dependent packages would be faster because doesn't need unpack and repack (into .txz) the whole texmf-tree.
I tried split the texmf-packages by their target (functionality) - I hope it's almost right (based on ArchLinux's splits: https://www.archlinux.org/packages/?q=texlive) but it's sure it isn't perfect.
Do you think isn't useful?

Looks good to me. If no one objects, I'll commit this after giving it a run in poudriere.

Too many bugs, postponing this for later.

Does that mean I can commit this myself, despite having commit bit for ports only?

Just out of curiosity, what's the rationale for this package split?

• Disable UIQT and SYNCTEX options, as these are causing compilation failures with GCC. Fix plist for DOCS and CONTEXT options.

• Refresh distfiles and host them on LOCAL, as they tend to change over time.

It seems, it still would be skipped by poudriere due to

The port is ready to be committed. I'll wait for couple of days in case someone wants to review it.

• Workaround LUA problem by using GCC to compile the port.

The LUA bug have not been fixed and upstream seem to be reluctant to fix it.

• Update to 2.9.7140.

Fixed upstream:

Another bump.

I've fixed the indentation and reuploaded the patch. @bcr, can you commit it?

I've uploaded the patch there: http://arrowd.name/doc.diff

Phab refused to upload the diff for non-UTF8 file, so I had to upload it as binary.

In D20825#450793, @bcr wrote:

OK, thanks. Do you want me to take care of the commit or are you doing it yourself?

Address comments.

In D20556#445213, @tcberner wrote:

Does for example x11-wm/hs-xmonad not also require a bump? It also has an EXECUTABLES=xmonad. In short, all hs-ports that define EXECUTABLES should be bumped, as their plist changes due to the new wrapper script: or did I misread the cabal.mk change?

Remove PLIST_FILES now that devel/hs-git-annex has pkg-plist.

In D20570#444950, @dg_syrec.org wrote:

In D20570#444928, @mat wrote:

Well, DEFAULT_VERSIONS is a user facing variable something users can set to choose the version they want to use. I have a feeling this is only for testing purposes, so it would probably be better to patch your local installation. Otherwise, people would end up setting go=go14 and complain things break.

Correct, go-devel is primarily for testing by port maintainers. The idea was to make this testing easier so most failures can be caught and fixed before the new lang/go version lands.

acme-client changes seem to be wrong.