You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
There is some drift/inconsistency/confusion in the various env vars we expose to control the R package build. One result of this is that we have crossbow jobs that are downloading nightly binaries instead of building from source. See discussion leading up to and including #39602 (comment).
Proposal:
FORCE_BUNDLED_BUILD should be deprecated in favor of LIBARROW_BUILD. I don't think we recommend it to people, it's more for our testing, and it's not named accurately anyway--it doesn't force a "bundled build", it routes you to nixlibs.R, where you'll either download a binary or get the bundled source build. Another way you can get configure to skip looking for libarrow on the system should be LIBARROW_BINARY equal to something other than false or unset--you're asking for a specific libarrow binary. (And today you can still get the same effect as FORCE_BUNDLED_BUILD by setting ARROW_USE_PKG_CONFIG=false and not having ARROW_HOME set.)
LIBARROW_BUILD=true should entail LIBARROW_BINARY=false. If LIBARROW_BUILD=true and LIBARROW_BINARY is something other than unset or false, raise an error.
Set download_libarrow_ok based on LIBARROW_BINARY=false, not LIBARROW_DOWNLOAD. Remove references to LIBARROW_DOWNLOAD. (Confirm if this download_libarrow_ok is necessary, or if LIBARROW_BINARY=false is already handled correctly.)
If we want something explicit about preventing downloading, that's TEST_OFFLINE_BUILD (as of GH-39601: [R] Don't download cmake when TEST_OFFLINE_BUILD=true #39602, it blocks all downloading). If that's a feature we want to advertise, we could rename that something that's not TEST_.
Describe the enhancement requested
There is some drift/inconsistency/confusion in the various env vars we expose to control the R package build. One result of this is that we have crossbow jobs that are downloading nightly binaries instead of building from source. See discussion leading up to and including #39602 (comment).
Proposal:
FORCE_BUNDLED_BUILD
should be deprecated in favor ofLIBARROW_BUILD
. I don't think we recommend it to people, it's more for our testing, and it's not named accurately anyway--it doesn't force a "bundled build", it routes you to nixlibs.R, where you'll either download a binary or get the bundled source build. Another way you can getconfigure
to skip looking for libarrow on the system should beLIBARROW_BINARY
equal to something other thanfalse
or unset--you're asking for a specific libarrow binary. (And today you can still get the same effect asFORCE_BUNDLED_BUILD
by settingARROW_USE_PKG_CONFIG=false
and not havingARROW_HOME
set.)LIBARROW_BUILD=true
should entailLIBARROW_BINARY=false
. IfLIBARROW_BUILD=true
andLIBARROW_BINARY
is something other than unset orfalse
, raise an error.download_libarrow_ok
based onLIBARROW_BINARY=false
, notLIBARROW_DOWNLOAD
. Remove references toLIBARROW_DOWNLOAD
. (Confirm if thisdownload_libarrow_ok
is necessary, or ifLIBARROW_BINARY=false
is already handled correctly.)TEST_OFFLINE_BUILD
(as of GH-39601: [R] Don't download cmake when TEST_OFFLINE_BUILD=true #39602, it blocks all downloading). If that's a feature we want to advertise, we could rename that something that's notTEST_
.ubuntu-r-only-r
job, and audit other jobs that use these env vars. (This may be covered by GH-39624: [R][CI] Add CMake to docker file and update envvars #39625)Component(s)
R
The text was updated successfully, but these errors were encountered: