rstudio-0.98.953 suddenly requires ALL the haskell [SOLVED?]

[ExecutorElassus]

The July 4th update to rstudio -- and another on the 5th -- added rstudio-0.98.953 and rstudio-0.98.935 to the tree, both of which have the following variable in RDEPEND:

[mackal]

Looks like a new feature depends on pandoc, which depends on all the Haskell. Best situation would be able to disable that feature, but I have no idea of it supports that :/

[gienah]

That is a pretty heavy dependency. Looking on the bright side, at least I bumped all
the haskell stuff in portage (which took days) just before the rstudio bump (which the
dev that bumped rstudio said took several hours, I'm not sure he has forgiven me
for nagging him to bump it).

You probably want to use the binary use flag for dev-lang/ghc.

When low level haskell packages like dev-haskell/text are bumped, it causes
lots of rebuilds of the other stuff. The haskell dependencies are complicated which
emerge sometimes has trouble trying to figure out what to do. If emerge
can figure it out, then it only builds the immediate reverse dependencies, then
its necessary to run:

[ExecutorElassus]

I'm hoping there's a USE variable to set so that it doesn't install any of them. I'm not too thrilled about maintaining a crapload of haskell stuff just so that rstudio can have some markup-conversion capability that I might never need. Here's hoping.

Cheers,

EE

[steveL]

[gienah]

The pandoc dep is a new feature, it looks like a hard dependency to me:

http://www.rstudio.com/products/rstudio/release-notes/

If you like you could try asking the rstudio developers if the pandoc
dependency could be made conditional:

https://support.rstudio.com/hc/communities/public/topics

However if pandoc was not available it appears to me that it would
break all of the R markdown functionality in rstudio. Unpacking
the rstudio tarball and running:

[ExecutorElassus]

There is, in fact, already a(n admittedly rudimentary) question about this on the forum:

https://support.rstudio.com/hc/communities/public/questions/200825318-pandoc-citeproc-dependency.

Here's hoping they address it.

[ExecutorElassus]

So, now there's this posting, which suggests that all that haskell stuff is just to install a more recent version of pandoc needed by rmarkdown.

That seems like a supremely wasteful way to ensure the most recent version of pandoc is on a system, especially when rstudio seems to ship with a version that works.

Is there a way to patch the ebuild to circumvent it trying to do what is pretty self-evidently needless and wasteful?

Cheers,

EE

[steveL]

There's always a way to patch an ebuild; that's the whole point of the format. No idea how effective it will be, nor do I use haskell or R. Best bet is to get help in #gentoo-dev-help and don't forget to learn #bash properly ;)

[ExecutorElassus]

hrm. Welp, the haskell stuff disappeared after a version bump to 0.9.1028. The reason for all the haskell in the first place was a hard dependency for a particular version of pandoc-citeproc; since the local installation didn't have it, rstudio tried to build it from source (thus all the haskell). Maybe with time the local installation caught up, or the ebuild now follows the patch I posted earlier.

Anyway, the problem went away, so, uh, [SOLVED], I guess?

Yay?

EE

[ExecutorElassus]

Ha ha, remember this thread, and how I thought the issue was solved?

It seems to have returned with 0.99.473-r1, in that it requires pandoc-citeproc again (which, along with ALL the haskell, is masked on my system for this very reason).

Guess I'll go ask #gentoo-dev-help?

Cheers,

EE

[gienah]

This thread discusses updating haskell packages:
https://forums.gentoo.org/viewtopic-t-1028484.html
In the 7th post is this suggested alternative to editing the sci-mathematics/rstudio ebuild to hack out the
dev-haskell/pandoc and dev-haskell/pandoc-citeproc dependencies.