Update Envoy to f84440dc4f95890f14e2e0686b07258f030b54b3#321
Update Envoy to f84440dc4f95890f14e2e0686b07258f030b54b3#321mum4k merged 20 commits intoenvoyproxy:masterfrom
Conversation
- Trace id generation broke. This was amended by directly using the new request id uid implementation. It would be nice to leverage the new request id extension capability later for this. - Amended calls to get counters by name - Synced .bazelrc - Amended test expectations for response size histograms which now report in with '.0' postfixed. Signed-off-by: Otto van der Schaaf <oschaaf@we-amp.com>
|
For this PR, ASAN failed for times in a row with: |
Signed-off-by: Otto van der Schaaf <oschaaf@we-amp.com>
Signed-off-by: Otto van der Schaaf <oschaaf@we-amp.com>
Signed-off-by: Otto van der Schaaf <oschaaf@we-amp.com>
Signed-off-by: Otto van der Schaaf <oschaaf@we-amp.com>
Signed-off-by: Otto van der Schaaf <oschaaf@we-amp.com>
Signed-off-by: Otto van der Schaaf <oschaaf@we-amp.com>
Signed-off-by: Otto van der Schaaf <oschaaf@we-amp.com>
Signed-off-by: Otto van der Schaaf <oschaaf@we-amp.com>
Signed-off-by: Otto van der Schaaf <oschaaf@we-amp.com>
Signed-off-by: Otto van der Schaaf <oschaaf@we-amp.com>
Signed-off-by: Otto van der Schaaf <oschaaf@we-amp.com>
Signed-off-by: Otto van der Schaaf <oschaaf@we-amp.com>
Signed-off-by: Otto van der Schaaf <oschaaf@we-amp.com>
Signed-off-by: Otto van der Schaaf <oschaaf@we-amp.com>
Signed-off-by: Otto van der Schaaf <oschaaf@we-amp.com>
Output formatting subtly changed, probably because of an updated dependency brought in by the new Envoy revision. Use an explicit formatting directive to amend and ensure that stays put. This also backs out the updated expectations, cleaning up the diff. Signed-off-by: Otto van der Schaaf <oschaaf@we-amp.com>
|
Readying this up for review. I plan to respin the tests a couple of times to make sure, but I think |
| echo "bazel TSAN debug build with tests" | ||
| echo "Building and testing envoy tests..." | ||
| cd "${SRCDIR}" | ||
| [ -z "$CIRCLECI" ] || export BAZEL_BUILD_OPTIONS="${BAZEL_TEST_OPTIONS} --local_ram_resources=12288" |
There was a problem hiding this comment.
Curios: what was the default value of the local_ram_resources flag?
There was a problem hiding this comment.
From https://docs.bazel.build/versions/master/user-manual.html:
By default Bazel will estimate amount of RAM and number of CPU cores directly
from system configuration.
| Envoy::Protobuf::util::TimeUtil::DurationToMicroseconds(percentile.duration())); | ||
| } else { | ||
| ss << fmt::format("{}.value: {}", percentile_prefix, percentile.raw_value()); | ||
| ss << fmt::format("{}.value: {:.{}f}", percentile_prefix, percentile.raw_value(), 0); |
There was a problem hiding this comment.
Not sure I follow this, what is the ", 0" for ?
There was a problem hiding this comment.
Basically the intent is to avoid trailing .0 by specifying zero precision. However, I changed this to use a static cast to an integer instead, which I'm hoping is more intuitive?
We want to avoid trailing '.0's in the output of percentile values. Effectively we want to represent integer values. So static cast the input to libfmt to int, instead of specifying a precision of 0. Signed-off-by: Otto van der Schaaf <oschaaf@we-amp.com>
Signed-off-by: Otto van der Schaaf <oschaaf@we-amp.com>
|
Ugh: ASAN ran out of memory again. It seems this now happens when building the Envoy It should be possible to scale back --jobs some more to resolve this. TSAN reported a different kind of failure. Looking into that, we're tracking that one over at #317. |
Signed-off-by: Otto van der Schaaf <oschaaf@we-amp.com>
mum4k
left a comment
There was a problem hiding this comment.
Thank you for the changes.
using the new request id uid implementation. It would be
nice to leverage the new request id extension capability
later for this.
to Envoy's dependency on
libfmt.Introduce an explicit formatting specifier in the affected output formatting code to address that.
revision. This affected both the build process as well as the testing process itself. Notably, the
python-based integration tests would get OOM-killed in asan runs.
Signed-off-by: Otto van der Schaaf oschaaf@we-amp.com