If NinjsReader keeps sending tool status messages after Close has been
called it can cause a concurrent map access when
CriticalPath.WriteToMetrics is called concurrently with
CriticalPath.FinishAction. Try harder to stop the NinjaReader goroutine
when NinjaReader.Close is called, even if the external ninja process has
not closed its FIFO or NinjaReader has not finished processing all the
messages after 5 seconds.
Bug: 286382228
Test: m nothing
Change-Id: I3e3dce601510e2dfb5ed82ca55bd11723fac7e70
The tags item in the trace data is arbitrary metadata added to `build`
steps by Soong (and my Kati via `.KATI_TAGS` target-specific
variables). Include this in the perfetto trace so we can analyze it.
Bug: http://b/259130368
Test: End to end test tracing cp time in dist targets
Change-Id: I85d33f579dc40dbae616b24cd4cb150d86262470
For the data to prioritize long running tasks in ninja, I added the
information about that.
Bug: 271526845
Test: build and check metrics
Change-Id: I76807f9b0d6a50d940b5f35dd6a9ef3cf693eec5
To improve build efficiency, the metrics for critical path and
parallelism ratio is necessary. That information has been included in
soong.log, so added it into metrics as well.
Bug: 271526845
Test: build and check if metrics pb has critical path info
Change-Id: I14e1a78c13d400b792d3b05df18604da48759ade
Major operations:
* Fix the go_package entries to be consistent, as the new tool
complains when it doesn't have a '/'.
* Regenerate with the new protoc-gen-go tool.
* github.com/golang/protobuf -> google.golang.org/protobuf
* proto.[Un]MarshalText -> prototext.[Un]Marshal
Change-Id: Ie1147bd2457fafb66ba555461b3bf14f0561a25d
The source tree will eventually be made ReadOnly, and recipes that write
directly to the source tree will fail. Use a pattern-match approach on
the results of stdout/stderr to provide hints to the user in such a
scenario.
If multiple patterns are found in raw output, print error hint
corresponding to first pattern match. first pattern match is chosen
since the failing function will be at the top of the stack, and hence
will be logged first
Test: Wrote a unit test to assert errorhint is added to output.
Wrote an integration test that writes to a file in the source tree
1. When source_tree is RO, the recipe fails and an error hint is printed
to stdout
2. When source tree is RW, the recipe succeeds and no error hint is
printed
Bug: 174726238
Change-Id: Id67b48f8094cdf8a571c239ae469d60464a1e89c
Propagate the subprocess stats from ninja to build.trace.gz.
There is too much data here to put into separate counters for each
statistic, so put them into args on each duration event for now.
We would like to track max RSS for each subprocess, but the Linux
kernel inherits the max RSS of the ninja process in each subprocess,
which sets a lower bound on the measurable max RSS to the size of
the ninja process. The ninja process is large due to the multi-GB
build.ninja files.
Bug: 170701554
Test: examine build.trace.gz
Change-Id: I8aaaafe627a57f1a500af098c097c6381c583ba5
This reverts commit 323dc60712.
Reason for revert: Possible cause of test instability
Bug: 170513220
Test: soong tests
Change-Id: Iee168e9fbb4210569e6cffcc23e60d111403abb8
Putting t.Parallel() in each test makes them run in parallel.
Additional t.Parallel() could be added to each subtest, although
that requires making a local copy of the loop variable for
table driven tests.
Test: m checkbuild
Change-Id: I5d9869ead441093f4d7c5757f2447385333a95a4
The stats output will now use the new "DEBUG" message type, which we can
always redirect to verbose.log.gz.
Test: m aprotoc (look in verbose.log.gz)
Change-Id: Ie1b58f12c008ff7d29f11ff7a9807488dba8a504
buld_error protobuf file is created the moment a build action failed.
This is for external system to detect immediately what caused the
build failure while monitoring the build_progress.pb file.
Bug: b/150401146
Test: * m nothing
* Built successfully aosp_arm-eng and checked build_error file
was not created.
* Modified a cpp file to cause a build failed. Ran m and build
failed. Checked immediately during the build that the
build_error file was created.
* m clean; m and during build, checked if build_error was
removed.
Change-Id: I778616ae8cf242b49ad263b79c93321959112caa
For successful builds, a blank build_error proto file is created.
This slows down the data processing part where it reads the build_error
file and there are no errors to be processed.
Bug: 142277430
Test: Did a successful build and checked if the build_error file
was not available. Changed a source file to make the build
failed and checked if the build_error file was available.
Change-Id: I300dffe32fb7f8bf984fa20ae368bdd02bf12992
Soong UI generates a proto file named build_progress.pb in $(OUT_DIR)
output directory that contains build action numbers (how many are executing,
finished and total) during the course of a build. This is for external
systems that invokes the Platform Build Systems and would like to know
the completion status.
Bug: b/150401146
Test: Wrote a bash script that continuously read the
build_progress.pb file and computed the build completed percentage
while building the aosp_arm-eng target. Compared the percentage between
the Soong output console and the one reported by the bash script.
Change-Id: I7c7347bc8e41958093892d8e2731c4f4169937dd
Soong UI will generate a new proto file in the out directory for
other systems to read the build completion status during a course
of a build. The proto file contains build action counters where
total_actions: represents the total build actions of a build (will
increase and may decrease during a course of the build)
finished_actions: represents the number of executed build actions.
finished_actions never decreases and finished_actions <= total_actions.
current_actions: represents the number of build actions being executed
and current_actions + finished_actions <= total_actions.
Bug: 150401146
Test: executed regen.sh script.
Change-Id: I04f4c07855723d0684b141d88cb3529ab5d9fccd
Use Printf instead of Println for formatted strings, and pass
the filename instead of the proto.
Test: cuj_tests
Change-Id: Id261b5a34304a0caa61faa1f3bbc388aacdd25a6
Keep a running map of the critical path to each edge as it
finishes. At the end of the build, find the edge with the
longest critical path and print the path to out/soong.log.
Test: critical_path_test.go
Change-Id: Ie01d26b068b768156b166bff00cdc3273e8124ca
Soong_ui produces a build_error raw file to $(OUT_DIR) where
the file contains a list of build action errors. Each build action
error represents an error raised by a command. The build_error file
is populated if there was a build error.
Bug: b/132969697
Test: Ran m for successful build. Introduced a broken build change,
ran m and verified using printproto that build_error was
generated successfully.
Change-Id: I690ca1778b4e56f144a3173ba1d16d8494c32c15
Write log output through StatusOutput so that the status implementation
can synchronize it with its own output.
Test: status_test.go
Change-Id: I917bdeeea4759a12b6b4aa6d6d86ee18a2771723
Wait for the ninja proto processing goroutine to notice the fifo
has closed and exit before continuing.
Test: m nothing
Change-Id: I8cf5f3b8bf6a91496c6d2bbbd3e811eb7f0c9d21
bufio.Scanner defaults to a relatively small max line length of
64kB. Kati output may have long lines, use a max length of 2MB
instead.
Fixes: 123590367
Test: m checkbuild
Change-Id: Ia869c5c2526621244f0065736e0d5fd9a088df06
This way we don't appear hung at:
No need to regenerate ninja file
Change-Id: I8dbdaa2c1b1c5a6a73187d0e6061f363b62e10c9
Fixes: 122251150
Test: m nothing
The idea is that we'd move the installation and packaging tasks over to
it, using data from Soong & the Kati reading Android.mk files.
This would allow us to make more fundamental changes about how we
package things without having to adjust makefiles throughout the tree.
Possible use cases:
* Moving some information from Soong's Android.mk output to a file read
by the packaging step may allow us to read the Android.mk files less
often, speeding up builds.
* Refactoring our current two-stage ASAN builds to run the Kati build
step twice, writing into different object directories, then have a
single packaging step that reads both outputs. Soong already has the
capability of writing out a single ninja file with all the asan
combinations.
* Running two build steps, one building the system-related modules
using a "generic" device configuration, and one building the vendor
modules using a specific device configuration. This could enforce a
GSI/mainline system vs vendor split in a single build invocation.
* If all installation is through this tool, it will be much easier to
track what should no longer be installed on an incremental build,
reducing the need for installclean.
* Changing PRODUCT_PACKAGES should be a much faster operation, which
means we could keep track of local additions to the images. Then
`mma` would be more persistent, instead of installing something once,
then never updating it again.
Eventually we plan on switching from Kati to something Go-based, but
this is a more incremental approach while we clean up everything else.
Currently, this just moves the dist-for-goal handling over to the
packaging step, so that we don't need to read Android.mk files when
DIST_DIR changes, or we switch between dist vs not.
Bug: 116968624
Bug: 117463001
Test: m nothing
Change-Id: Idec5ac6f7c7475397ba0fb65bd3785128a7517df
The extra `defer os.Remove(fifo)` was sometimes racing with the next
instance of ninja, removing the file in between when it was re-created
and used. Since we're always removing the file before creating it, it's
safe to just remove that.
The error message for this failure wasn't all that good either, so move
so use the status Print/Error calls inside the goroutine instead of the
logging ones.
Test: `build/soong/build_test.bash -only-soong` repeatedly
Change-Id: Icfeb6b68802093bd3a07d3e46046ef7d1a89d4a1
This adds a new status package that merges the running of "actions"
(ninja calls them edges) of multiple tools into one view of the current
state, and gives that to a number of different outputs.
For inputs:
Kati's output parser has been rewritten (and moved) to map onto the
StartAction/FinishAction API. A byproduct of this is that the build
servers should be able to extract errors from Kati better, since they
look like the errors that Ninja used to write.
Ninja is no longer directly connected to the terminal, but its output is
read via the protobuf frontend API, so it's just another tool whose
output becomes merged together.
multiproduct_kati loses its custom status routines, and uses the common
one instead.
For outputs:
The primary output is the ui/terminal.Status type, which along with
ui/terminal.Writer now controls everything about the terminal output.
Today, this doesn't really change any behaviors, but having all terminal
output going through here allows a more complicated (multi-line / full
window) status display in the future.
The tracer acts as an output of the status package, tracing all the
action start / finish events. This replaces reading the .ninja_log file,
so it now properly handles multiple output files from a single action.
A new rotated log file (out/error.log, or out/dist/logs/error.log) just
contains a description of all of the errors that happened during the
current build.
Another new compressed and rotated log file (out/verbose.log.gz, or
out/dist/logs/verbose.log.gz) contains the full verbose (showcommands)
log of every execution run by the build. Since this is now written on
every build, the showcommands argument is now ignored -- if you want to
get the commands run, look at the log file after the build.
Test: m
Test: <built-in tests>
Test: NINJA_ARGS="-t list" m
Test: check the build.trace.gz
Test: check the new log files
Change-Id: If1d8994890d43ef68f65aa10ddd8e6e06dc7013a
