Correct Windows build defects noted with recent version updates#10469
Correct Windows build defects noted with recent version updates#10469mattklein123 merged 3 commits intoenvoyproxy:masterfrom greenhouse-org:review-externals-libcmt
Conversation
- Pick up curl-7_69_1 - Patch curl to revert change to cmake_minimum_required - latest version of curl added a maximum version of cmake 3.16 - Pick up cares-1_16_0 - Toggle crosstool build to avoid possibly injecting wrongly assumed flags (/MD or -MD dynamic linkage override of our /MT static build) Signed-off-by: Sunjay Bhatia <[email protected]> Signed-off-by: William A Rowe Jr <[email protected]> Co-authored-by: Sunjay Bhatia <[email protected]> Co-authored-by: William A Rowe Jr <[email protected]>
Drop the changes specifing crosstool compilation, as this broke OS/X (and introduced the format failure) Signed-off-by: Sunjay Bhatia <[email protected]> Signed-off-by: William A Rowe Jr <[email protected]> Co-authored-by: Sunjay Bhatia <[email protected]> Co-authored-by: William A Rowe Jr <[email protected]>
The 1.16.0 release is unusable by Envoy, according to the linux asan and tsan CI builds. Co-authored-by: Sunjay Bhatia <[email protected]> Co-authored-by: William A Rowe Jr <[email protected]> Signed-off-by: Sunjay Bhatia <[email protected]> Signed-off-by: William A Rowe Jr <[email protected]>
|
Assigning to @junr03 to potentially help sort out the c-ares issues. |
|
From the ASAN failure in https://dev.azure.com/cncf/4684fb3d-0389-4e0b-8251-221942316e06/_apis/build/builds/35151/logs/68 It looks to me that this is a bug in c-ares: else if (status == ARES_EDESTRUCTION)
{
end_hquery(hquery, status);
}
if (!hquery->remaining)
|
|
Some failing macos tests from the previous bump PR (see: this build) and this one (see: this build)
|
|
@mattklein123 this patch makes no changes to affecting quiche or the ActiveQuicListenerTest, as OS/X reports. We should move ahead with this patch to unblock the curl update. |
|
This PR didn't end up updating c-ares from the looks of it @wrowe : and change to c-ares patch as per #10433 (comment) |
|
You are correct. We began by updating c-ares in the same lot as curl. However, as noted above, c-ares 1.16 introduced new regressions that must be addressed before it is ready for master. |
|
(From the comments above, it appears c-ares regressions are upstream code defects, not envoy's use of the library.) |
Description:
Picks up from #10433
windows build regression...
flags (/MD or -MD dynamic linkage override of our /MT static build)
Risk Level: Low
Testing: Local on MSVC
Docs Changes: n/a
Release Notes: n/a