Simplified detection of macOS for install_name_tool#146
Conversation
|
In addition to the uname checks, I believe need need a check for "Are we linking against the downloaded library". The problem with conda is we don't link against the downloaded library but a system install and that is where the use of |
So just add a I was leaning towards moving the "shall we do this?" question into |
|
And/or we add a third logical condition that will be TRUE for normal builds and FALSE for Conda. Filled in by And/or we use an env.var for conda and if set skip this. How would I detect a conda build? |
I'm much more in favor of a generic solution. Conda is just one env where we'll have a preexisting library. Seems like development would be another common use case were we want to avoid trying to run |
|
We already cover preexisting libraries, and allow building, and only download where needed. So we already are more general: edd@rob:~/git/tiledb-r(master)$ ./configure --help
`configure' configures TileDB-R 0.7.1 to adapt to many kinds of systems.
[...]
--enable-building build TileDB library locally instead of downloading
prebuilt
[...]
--with-tiledb=PREFIX path to installed built of TileDB and of course if the library is in a standard location it "just works" without any options. Which is, in essence, all my builds. |
with a tip of the hat to to data.table and an additional file-present test
25903c5 to
ca5d0a5
Compare
|
Helpful discussion on and off-repo --- I just updated the test to check for (both) files before potentially hitting them with |
The previous addition to protect use of
install_name_toolupset Conda, so it was changed.Which upsets
R CMD checkbecause it usedifneqwhich, according to the gospel of CRAN, is non-portable.So I changed it again, borrowing the construct from
data.tablewhich, as I checked, has a conda build so the new double test appears to hold water. Actual call unchanged, all we loose is the "non-portable if" and an echo statement we do not really need.