non-cross-compile ci builds should not hardcode an architecture, so they can be run on any ci system
ci: Run non-cross-compile builds natively #17367
pull MarcoFalke wants to merge 3 commits into bitcoin:master from MarcoFalke:1911-ciNative changing 9 files +13 −14-
MarcoFalke commented at 7:04 PM on November 4, 2019: member
- MarcoFalke added the label Tests on Nov 4, 2019
-
ci: Run CI_WAIT only on travis fa56bcbb01
-
ci: Run non-cross-compile builds natively fa8a60bce9
- MarcoFalke force-pushed on Nov 4, 2019
-
laanwj commented at 1:54 PM on November 6, 2019: member
ACK fa8a60bce9a0b407f8fbed5409fccaad1204b7e7
-
ci: Guess the native host when not cross compiling faf757a125
-
in ci/test/00_setup_env.sh:27 in fa8a60bce9 outdated
20 | @@ -21,6 +21,9 @@ echo "Fallback to default values in env (if not yet set)" 21 | export MAKEJOBS=${MAKEJOBS:--j4} 22 | # A folder for the ci system to put temporary files (ccache, datadirs for tests, ...) 23 | export BASE_SCRATCH_DIR=${BASE_SCRATCH_DIR:-$BASE_ROOT_DIR/ci/scratch/} 24 | +# What host to compile for. See also ./depends/README.md 25 | +# Tests that need cross-compilation export the appropriate HOST. 26 | +# Tests that run natively, do not set a HOST, but we assume x86_64 here if the calling environment did not set a HOST 27 | export HOST=${HOST:-x86_64-unknown-linux-gnu}
laanwj commented at 1:55 PM on November 6, 2019:What is this host define used for? Could this be auto-detected (if not explicitly set for cross-build) instead of assumed?
MarcoFalke commented at 2:18 PM on November 6, 2019:The host is passed down to depends and then used there to cross compile. Also, it is used to name the distdir folder.
MarcoFalke commented at 2:18 PM on November 6, 2019:So if you are not using depends, the host is only used to name the distdir folder.
MarcoFalke commented at 2:19 PM on November 6, 2019:So I am not sure if it makes sense to add auto-detect code to pick the right temporary name
laanwj commented at 2:21 PM on November 6, 2019:It does look too much like "hardcoding an architecture" to me, which the point of this PR is to avoid. If it's only used to name a folder, can't we name it something architecture-independent? (like "native") (I'd dread to ever see ARM objects in a x86_64-unknown-linux-gnu directory and it not being a bug :smile: )
MarcoFalke commented at 5:21 PM on November 7, 2019:Fixed
MarcoFalke force-pushed on Nov 7, 2019MarcoFalke commented at 8:27 PM on November 7, 2019: memberlaanwj commented at 9:06 PM on November 8, 2019: memberre-ACK faf757a1254a57aec7868ea6dffd062623e40756
laanwj referenced this in commit 9e0aabe50c on Nov 8, 2019laanwj merged this on Nov 8, 2019laanwj closed this on Nov 8, 2019sidhujag referenced this in commit c6375a109d on Nov 9, 2019MarcoFalke deleted the branch on Nov 11, 2019sidhujag referenced this in commit 5101cbbab1 on Nov 10, 2020MarcoFalke locked this on Dec 16, 2021ContributorsLabels
This is a metadata mirror of the GitHub repository bitcoin/bitcoin. This site is not affiliated with GitHub. Content is generated from a GitHub metadata backup.
generated: 2026-04-17 06:14 UTC
More mirrored repositories can be found on mirror.b10c.me