Chris Wolfe has posted comments on this change.
Change subject: Add USE=ninja support to chromeos-chrome.
......................................................................
Patch Set 2: (8 inline comments)
Quick drive-by with the things that jumped out at me. Don't worry about ticking them off now.
....................................................
File chromeos-base/chromeos-chrome/chromeos-chrome-9999.ebuild
Line 129: if ! use ninja; then
I'd really prefer we not do this -- moving the output files around is a good way to confuse people. For any workflow other than sshfs-mounting the build directory using /var/cache should be faster (the ebuild install can then use hard-links).
Line 210: BUILD_TOOL_FLAGS=(
Is it always safe to pass these flags to the make generator? I think we're probably okay, but it makes me paranoid.
Line 522: export GOMANINJA=${GOMANINJA:-/home/$(whoami)/depot_tools/ninja-linux6}
Unused?
Line 523: export ENINJA="${ENINJA:-/home/$(whoami)/depot_tools/ninja-linux64}"
Prefer NINJA, so nobody will expect eutil-like behavior.
Line 701: src_configure() {
Feels like "do runhooks in src_configure" should be landed as a separate refactoring. I agree with it (and proposed it a while ago), but it needs to be tested against the ebuild-based tools.
Line 716: export CC_host=gcc
Believe these should be $(tc-getBUILD_CC) and similar.
Line 778: $(use chrome_internal && default_extensions) \
missing echo
These use&&echo blocks would be simpler in EAPI=4 with usex, but wouldn't do that without input from one of the Portage gurus.
Line 956: if use ninja; then
Getting back to the same cache dir should also avoid this.
Gerrit-MessageType: comment
Gerrit-Change-Id: I402e9252d79a7311ecc43f9273ea3f967509d59e
Gerrit-PatchSet: 2
Gerrit-Project: chromiumos/overlays/chromiumos-overlay
Gerrit-Branch: master
Gerrit-Owner: Taiju Tsuiki <
tz...@chromium.org>
Gerrit-Reviewer: Chris Wolfe <
cwo...@chromium.org>
Gerrit-Reviewer: Taiju Tsuiki <
tz...@chromium.org>