Issue metadata
Sign in to add a comment
|
Mac builders for BoringSSL are missing Xcode |
||||||||||||||||||||||
Issue descriptionSeeing this on CI and try bots. https://ci.chromium.org/p/boringssl/builders/luci.boringssl.try/ios64_compile/b8934128890223096144 https://ci.chromium.org/p/boringssl/builders/luci.boringssl.ci/mac/b8934215559584825008 This is a different failure from issue #888687 so filing this separately. With both failing, we are entirely without macOS/iOS coverage in BoringSSL right now.
,
Sep 28
We do not provide Xcode via the system image anymore. You need to be utilizing the hermetic Xcode toolchain
,
Sep 28
Do you have instructions on how that is set up? BoringSSL is a standalone project, so we don't pick up Chromium's bits for free. Also, was this announced anywhere? Would have been nice to have a smoother transition.
,
Sep 28
Over to troopers for assistance. Troopers - could someone point davidben@ to the right spot on how to include the hermetic Xcode bits?
,
Sep 28
Your recipe should be updated to do something similar to the code at https://codesearch.chromium.org/chromium/build/scripts/slave/recipe_modules/chromium/api.py?rcl=89f534b2289b253beb4d742ed039a3a557eefc89&l=820
,
Sep 28
Alright, I'll take a look at that. Where should we subscribe for these sorts of changes, so we can prepare ahead of time for the next one?
,
Sep 29
Routing to ICC to answer davidben@ question. I personally don't have a good answer :(
,
Oct 1
Perhaps sergeyberezin or jbudorick@ could give some pointers on how to fix?
,
Oct 1
The best way is to install Xcode per build - similar to https://crbug.com/884015#c38 (actual CL https://crrev.com/c/1241857). Basically, add depot_tools/osx_sdk recipe module, and run your build / tests in its context. Also add the appropriate service accounts to https://chrome-infra-auth.appspot.com/auth/groups/ios-xcode-readers (or go/bugatrooper for assistance - non-troopers are not likely to have enough permissions).
,
Oct 1
Ooh, thanks! That looks quite straightforward!
,
Oct 1
Re ACLs, we're just using luci.flex.ci. I'm not sure what service account that corresponds to.
,
Oct 1
The following revision refers to this bug: https://boringssl.googlesource.com/boringssl/+/cbc3e076fc0eb3d21870bbf83e0ced1a0f01a6e3 commit cbc3e076fc0eb3d21870bbf83e0ced1a0f01a6e3 Author: David Benjamin <davidben@google.com> Date: Mon Oct 01 17:41:54 2018 Take iOS builders out of the CQ rotation too. Bug: chromium:890351 Change-Id: Ia11b2b97f25d0c37e491641db6c48aa37c03de30 Reviewed-on: https://boringssl-review.googlesource.com/32224 Reviewed-by: Adam Langley <agl@google.com> [modify] https://crrev.com/cbc3e076fc0eb3d21870bbf83e0ced1a0f01a6e3/infra/config/cq.cfg
,
Oct 1
Re #c3: unfortunately, this change was mostly confined to Chromium on Mac and iOS, and was announced on their lists only. It was only with the migration to LUCI and consolidation of machine pools we realized that so many other teams are implicitly relying on system Xcode. It's a good idea to announce it wider now, but I'm also not sure what a good list would be. Maybe infra-dev@chromium.org?
,
Oct 1
Looking at the swarming tasks for the builds referenced in #0, the service accounts are: boringssl-try-builder@chops-service-accounts.iam.gserviceaccount.com boringssl-ci-builder@chops-service-accounts.iam.gserviceaccount.com I added those to the group.
,
Oct 1
Thanks! https://chromium-review.googlesource.com/c/chromium/tools/build/+/1255546 https://boringssl-review.googlesource.com/c/boringssl/+/32244
,
Oct 1
The following revision refers to this bug: https://boringssl.googlesource.com/boringssl/+/198a703f6bd7782112ab62942a48c7c77e10a26b commit 198a703f6bd7782112ab62942a48c7c77e10a26b Author: David Benjamin <davidben@google.com> Date: Mon Oct 01 22:39:53 2018 Add an osx_sdk cache to BoringSSL builders. This will carry the hermetic toolchain for Mac builders. It was suggested that using mixins rather than adding empty caches to each builder was preferable, so this also makes the try builders use mixins. This CL adds a CPU selector to luci.boringssl.ci to match luci.boringssl.try. It's a no-op, but we may as well note that we do indeed want an x86_64 machine. It also aligns the macOS versions a bit. The try jobs were using a mix of 10.9 and 10.12 because, at the time, iOS required 10.12 but 10.9 was more plentiful. Now 10.13 is more plentiful. Using the mixin is probably better so we can update it all at once. (There's something to be said for running the tests on older versions, but we haven't been doing that historically. That kind of coverage is probably more efficient to do on the embedder side like Chromium.) There's a similar version disparity on Windows between Win7 and Win10. I've left those alone in this CL and added a "win7" mixin. https://chromium-review.googlesource.com/c/chromium/tools/build/+/1255546 is the recipe-side change. Bug: chromium:890351 Change-Id: I11bbddaa513ec5b48dacb2626bd7b282473d4a43 Reviewed-on: https://boringssl-review.googlesource.com/32244 Reviewed-by: David Benjamin <davidben@google.com> Reviewed-by: Sergey Berezin <sergeyberezin@chromium.org> Commit-Queue: David Benjamin <davidben@google.com> CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org> [modify] https://crrev.com/198a703f6bd7782112ab62942a48c7c77e10a26b/cr-buildbucket.cfg
,
Oct 1
The following revision refers to this bug: https://chromium.googlesource.com/chromium/tools/build/+/a21ffa47db966c59f3e98c96ff5e80bf857041b5 commit a21ffa47db966c59f3e98c96ff5e80bf857041b5 Author: David Benjamin <davidben@google.com> Date: Mon Oct 01 22:57:18 2018 boringssl: Pull in the hermetic toolchain on Mac. https://boringssl-review.googlesource.com/c/boringssl/+/32244 adds the cache on the BoringSSL side. Bug: 890351 Change-Id: If5e5069d96b30cc3338becb5a8f30519d7c965fc Reviewed-on: https://chromium-review.googlesource.com/1255546 Commit-Queue: David Benjamin <davidben@chromium.org> Reviewed-by: Sergey Berezin <sergeyberezin@chromium.org> [modify] https://crrev.com/a21ffa47db966c59f3e98c96ff5e80bf857041b5/scripts/slave/recipes/boringssl.expected/mac_small.json [modify] https://crrev.com/a21ffa47db966c59f3e98c96ff5e80bf857041b5/scripts/slave/README.recipes.md [modify] https://crrev.com/a21ffa47db966c59f3e98c96ff5e80bf857041b5/scripts/slave/recipes/boringssl.expected/mac_rel.json [modify] https://crrev.com/a21ffa47db966c59f3e98c96ff5e80bf857041b5/scripts/slave/recipes/boringssl.expected/ios64_compile.json [modify] https://crrev.com/a21ffa47db966c59f3e98c96ff5e80bf857041b5/scripts/slave/recipes/boringssl.py [modify] https://crrev.com/a21ffa47db966c59f3e98c96ff5e80bf857041b5/scripts/slave/recipes/boringssl.expected/mac.json [modify] https://crrev.com/a21ffa47db966c59f3e98c96ff5e80bf857041b5/scripts/slave/recipes/boringssl.expected/ios_compile.json
,
Oct 1
The following revision refers to this bug: https://boringssl.googlesource.com/boringssl/+/9e97c022e6459035625b842451fd8582fe022dcc commit 9e97c022e6459035625b842451fd8582fe022dcc Author: David Benjamin <davidben@google.com> Date: Mon Oct 01 23:31:45 2018 Bring Mac and iOS builders back to the CQ. The vpython issue appears to have gone away and hermetic Xcode sorted out the other problem. Bug: chromium:888687 , chromium:890351 Change-Id: I9da893b7f21f0bc7c03e1e70c0e3e86f9720cec1 Reviewed-on: https://boringssl-review.googlesource.com/32304 Reviewed-by: Adam Langley <agl@google.com> Commit-Queue: David Benjamin <davidben@google.com> CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org> [modify] https://crrev.com/9e97c022e6459035625b842451fd8582fe022dcc/infra/config/cq.cfg
,
Oct 1
,
Oct 2
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/6bd7dab6856e45a31b5b234b79fcb6ddfc56754f commit 6bd7dab6856e45a31b5b234b79fcb6ddfc56754f Author: David Benjamin <davidben@chromium.org> Date: Tue Oct 02 21:47:18 2018 Roll src/third_party/boringssl/src 13fd62744..ce00828c8 https://boringssl.googlesource.com/boringssl/+log/13fd627449cefbae7576d4b145cb24fac303fc7d..ce00828c89df3d4c40de7d715b1a032eb03c525c The following commits have Chromium bugs associated: 9e97c022e Bring Mac and iOS builders back to the CQ. 7c3ce519e Actually disable RandTest.Fork on iOS. 52483994c Mostly fix undefined casts around STACK_OF's comparator. fb4e2e0f0 Fix undefined casts in sk_*_pop_free and sk_*_deep_copy. cbc3e076f Take iOS builders out of the CQ rotation too. 792c1dc43 Rewrite PEM_X509_INFO_read_bio. 419144adc Fix undefined function pointer casts in {d2i,i2d}_Foo_{bio,fp} 217bfd3c9 Fix undefined function pointer casts in IMPLEMENT_PEM_*. I've also updated the roll_boringssl.py script to print the above message and the below bug list, so we don't need to manually attach things to rolls. Bug: 785442 , 888687 , 890115, 890351 Change-Id: If08496fcff132e71052a89c60aefb609d28ad3c7 Reviewed-on: https://chromium-review.googlesource.com/c/1257561 Commit-Queue: Steven Valdez <svaldez@chromium.org> Reviewed-by: Steven Valdez <svaldez@chromium.org> Cr-Commit-Position: refs/heads/master@{#595994} [modify] https://crrev.com/6bd7dab6856e45a31b5b234b79fcb6ddfc56754f/DEPS [modify] https://crrev.com/6bd7dab6856e45a31b5b234b79fcb6ddfc56754f/third_party/boringssl/roll_boringssl.py |
|||||||||||||||||||||||
►
Sign in to add a comment |
|||||||||||||||||||||||
Comment 1 by davidben@chromium.org
, Sep 28