| Age | Commit message (Collapse) | Author |
|---|
|
A few of the unittests fail when running on ARM. Android has already
fixed those problems and protected them with __ANDROID__. We can just
start using them in Chrome OS too.
BUG=b:165005216
TEST=FEATURES=test emerge-gale update_engine
Change-Id: I7f8820c2ba81f23144257d78aa6c0c599f645e6a
Reviewed-on: https://chromium-review.googlesource.com/c/aosp/platform/system/update_engine/+/2460235
Tested-by: Amin Hassani <ahassani@chromium.org>
Reviewed-by: Jae Hoon Kim <kimjae@chromium.org>
Commit-Queue: Amin Hassani <ahassani@chromium.org>
|
|
MessageLoop::current()->WatchFileDescriptor is deprecated. And UE should
remove usages of it.
Test: mma && unittest
Change-Id: Ib1ef2e6b6a38ad2a8d07b78bcd72fdb3b7f82226
|
|
It's a merge from chrome OS with some reverts.
1. the fd watcher change, because the libbrillo version isn't
compatible in aosp.
commit 6955bcc4ffe4cc9d62a88186b9a7e75d095a7897
commit 493fecb3f48c8478fd3ef244d631d857730dd14d
2. two libcurl unittest. Because the RunOnce() of the fake message
loop seems to have different behavior in aosp.
commit d3d84218cafbc1a95e7d6bbb775b495d1bebf4d2
Put preprocessor guards to use the old code in aosp. And we can
switch to the new code in the other path after adopting the new
libbrillo & libchrome.
Test: unit tests pass, apply an OTA
Change-Id: Id613599834b0f44f92841dbeae6303601db5490d
|
|
Whenever we are dealing with some problem that require defining error
codes and sending UMA metrics, we need to define a new ErrorCode enum
value. These error codes then will be used to send which UMA metric
value to send. Some of the UMA metrics seems to have circumvent this
process by introducing their own error codes without adding them to the
global list of error codes.
This CL introduces three new error codes:
- kInternalLibCurlError
- kUnresolvedHostError
- kUnresolvedHostRecovered (Technically not an error code, but fits the
description and use case of it.)
That are then translated to the UMA metric values we send for
DownloadErrorCode.
In addition, this CL moves the responsibility of sending these UMA
metrics from LibCurlHttpFetcher to OmahaRequestAction which is the more
correct place to send it because that's where the operations are
completed (success or failure) and we can safely decide on the value of
UMA without risking to send overlapping or duplicated metrics.
For example, previously we send kInternalLibCurlError in conjunction
with the kUnresolvedHostError. But doing this can hide the fact that
these two error codes might be related and caused by the same underlying
issue.
Same goes for kUnresolvedHostError and kUnresolvedHosRecovered. If we
send both these metrics at the same time, then we need to subtract the
number of kUnresolvedHosRecovered from kUnresolvedHostError to figure out
the number of unresolved host errors that did not recover. By
exclusively sending one or another. We can see exactly how many are
recovered and how many did not. Although this might change the meaning
of kUnresolvedHostError metric, in the long term it will not be an issue
as all the results will converge to the new behavior.
The enum.xml (chrome) is updated in crrev.com/c/1774101
BUG=None
TEST=cros_workon_make --board=amd64-generic --test --noreconf update_engine
Change-Id: I3c7bb5f6159a0bc3a37d55666572b9cd6730f3cb
Reviewed-on: https://chromium-review.googlesource.com/c/aosp/platform/system/update_engine/+/1759544
Reviewed-by: Nicolas Norvez <norvez@chromium.org>
Tested-by: Amin Hassani <ahassani@chromium.org>
Commit-Queue: Amin Hassani <ahassani@chromium.org>
|
|
libcurl error
Based on https://curl.haxx.se/docs/todo.html#updated_DNS_server_while_running:
"If /etc/resolv.conf gets updated while a program using libcurl is running, it
may cause name resolves to fail unless res_init() is called. We should
consider calling res_init() + retry once unconditionally on all name resolve
failures to mitigate against this."
This CL added following behavior:
On libcurl returns CURLE_COULDNT_RESOLVE_HOST error code:
1. we increase the max retry count by 1 for the first time it happens in the
lifetime of an LibcurlHttpFetcher object.
2. we call res_init unconditionally.
We also add UMA metrics to measure whether calling res_init helps
mitigate the unresolved host problem. WIP CL: https://chromium-review.googlesource.com/c/chromium/src/+/1698722
BUG=chromium:982813
TEST=FEATURES="test" emerge-kefka update_engine, tested on a device
Change-Id: Ia894eae93b3a0adbac1a831e657b75cba835dfa0
|
|
In order for Omaha to correlate Omaha Client requests with the actual
binary download, the session ID must be attached to the HTTP header of
the binary download in the X-Goog-Update-SessionId.
Also, remove the HTTP header of X-Goog-Update-SessionId added into the
Omaha requests.
BUG=chromium:940515
TEST=unittests # new unittests
Change-Id: I0759562f2d1c8c003064ad976ca1ae6ce039b960
|
