This PR adds a simple daily job that runs the performance test on a
chosen "baseline" commit and then runs the same benchmark on latest
master. This should allow us to track overall performance improvements.
CHANGELOG_BEGIN
CHANGELOG_END
* docs: publish daml-cheat-sheet on docs.daml.com
* separate cheat sheet rule
* Fixed `@daml-cheat-sheet`
* Uses `_config.yml` to determine the source directory root.
* Uses `tar h` to resolve symbolic links. Otherwise the tarball just
contains symbolic links to the execroot.
* Uses flags to make `tar` and `gzip` reproducible, i.e. avoid
timestamps and the like.
* cleanup
Co-authored-by: Andreas Herrmann <andreas.herrmann@tweag.io>
* Diff with messaging feature and some noise manually removed
* Bazel target to use patch file in other build targets
* Patch file as data dep for integration tests
* Attempt to patch and test messaging feature in create-daml-app test
changelog_begin
changelog_end
* Use exports_files instead of filegroup
* Remove file existence checks that don't make sense
* Add patch to dev_env and reference it from integration tests
* Include patch on windows for later
* Set up yarn env again after codegen
* Restore file check
* Fix typo in comment on util function
* Add Tasty steps to make process explicit
* Use messaging patch for code snippets in GSG
* Use messaging code from template instead of copy
* Remove copied message code
* Refactor script to copy template code with messaging patch
* No need to retry second yarn install (only local deps should be updated)
changelog_begin
changelog_end
For now this is only used for the daml-helper tests. I’ll shuffle
things around and use it for all tests in a separate PR.
* Include Bazel patch to mark tests as exclusive
This should hopefully avoid rerunning the conformance tests as often
as we do now. While this patch is not applied on Windows (since we get
it from nix), this is not really an issue since most of the exclusive
tests (in particular, all conformance tests) are disabled on Windows
anyway.
I’ve tested this locally accross a couple of runs and I get the
caching I want and looking at the code in the patch, the change looks
very reasonable. I somewhat wonder if it just broke internally at
google because they marked tests as exclusive that should have gotten
no-cache.
changelog_begin
changelog_end
* Disable caching for canton
This reduces the number of GHCs to 2 on Linux (regular and DWARF) and
1 on macOS. Given that each derivation is > 1 GB this should hopefully
help a bit.
changelog_begin
changelog_end
This disables the PDF docs builds on MacOS on CI (they are still built
locally by default) and removes them from the Nix closure by
introducing a separate ci-cached attribute that filters out texlive.
Since we built `nix-build nix -A tools -A cached` on CI, I’ve also
removed all the Tex stuff from tools which only means that it ends up
in PATH which nobody seems to care about.
changelog_begin
changelog_end
* Remove unused npm_nix
* Use node_nix directly on Unix
rules_nodejs requires a vendored node to contain the node distribution
in a subdirectory of the provided external repository. With the default
node Nix package this is not the case. The nodejs_dev_env workspace used
an intermediate repository rule to fulfill these requirements.
Occasionally, this caused flakiness on CI.
This change fulfills this requirement on the Nix side, so that Bazel can
consume the provided Nix derivation directly.
CHANGELOG_BEGIN
CHANGELOG_END
* Avoid collision with dev-env tools like license-checker
* buildifier-fix
* Add comment motivating use of `@node_nix` on Unix.
* Document npm-cli patch
Co-authored-by: Andreas Herrmann <andreash87@gmx.ch>
* Get grpc from nix on unix
The one from Bazel seems to cause linking issues when trying to run
things in GHCi. I’ve spent some time trying to use rules_foreign_cc to
build gRPC using CMake but decided that for my own sanity it’s better
to not pursue that further.
* Address review comments
* Add missing module load
* Cleanup GHCI_SCRIPT
* use the correct file ending on macos
* Import is_linux
* Switch back to grpc-1.23
The newer version seems to cause issues in combination with the java libraries.
* Try to fix package_app on macos
* more debugging
* Maybe this is not necessary, we will never know
* linkers are the worst
* Remove debugging output again
* readd rpaths
* treat libdispatch specially
* remove hack
* more fooling around
* lalala
* Fix pip and pipenv
Since the latest nixpkgs update pip is no longer contained in the python
derivation but instead in its own `python37Packages.pip` derivation.
Additionally the pipenv version in nixpkgs is incompatible with the pip
version, as pipenv expects a module to be present which is only
available in later versions. This change overrides pip with a version
update to fix this issue.
Upstream nixpkgs has already updated pip accordingly
* Update nixpkgs
The versions of `pipenv` and `pip` contained in `nixpkgs` where
previously incompatible. They are compatible in the new `nixpkgs`
revision.
Just overriding the pip version caused too many uncached rebuilds and
timed out on CI.
* Update nixpkgs further
To incorporate
efce3c1367
* Try to build fewer GHCs
* nix: drop sass override
nixpkgs-unstable is on the same page.
* nix: use upstream chromedriver
* nix: remove pex override
* nix: drop pypi2nix override
* Fix bazel-watcher
* try upgrading rules go
* Remove overrides for ibazel
* Upgrade rules_go
* Remove references to sass overrides
* ledger-api-test-tool-on-canton: Run Canton.
* ledger-api-test-tool-on-canton: Run the SemanticTests against Canton.
This was _so_ much work.
These will probably be flaky, so can't merge them in until we fix the
underlying issues with the tests around multi-participant allocation.
* ledger-api-test-tool: Wait for parties to arrive on all participants.
Thanks to Canton for exacerbating this bug.
Can't turn on all the tests on Canton yet as there are other spurious
failures. Will investigate next.
* ledger-api-test-tool: Move all contract key tests to ContractKeys.
If a ledger doesn't support ContractKeys, they need to be able to turn
these tests off.
* ledger-api-test-tool: If a test name is misspelled, fail immediately.
I keep getting names slightly wrong in the `--exclude` arguments and
wondering why it didn't work.
* client_server: Revert client_server_test.bzl.
Turns out I don't know what I'm doing. ¯\_(ツ)_/¯
* sandbox: Don't call `.toString` on an array. Doesn't do much.
* dev-env: Don't untar netcat automatically.
It needs to be installed with Pacman.
* Use cc_wrapper in ghcide test
The cc_wrapper is needed to find library dependencies.
* Update rules_haskell
* update rules_nixpkgs
Fixes issue with `nixpkgs_local_repository` and Nix `import`.
* Shorten RULES_HASKELL_EXEC_ROOT
Cabal package library paths are longer and overflow the command-line
length when calling `cc` on MacOS for template Haskell dependencies in
the IDE tests. Shortening the `RULES_HASKELL_EXEC_ROOT` prefix for each
`(LD_)LIBRARY_PATH` entry works around the issue.
* bazel: 0.28.1 --> 1.1.0
* bazel-watcher sha256
* Fix missing line in patch
* proto_source_root --> strip_import_prefix
See https://github.com/bazelbuild/bazel/issues/7153 for details.
* Update rules_nixpkgs
Required to avoid errors of the form
```
ERROR: An error occurred during the fetch of repository 'node_nix':
parameter 'sep' may not be specified by name, for call to method split(sep, maxsplit = None) of 'string'
```
and
```
ERROR: An error occurred during the fetch of repository 'node_nix':
Traceback (most recent call last):
File "/private/var/tmp/_bazel_runner/17d2b3954f1c6dcf5414d5453467df9a/external/io_tweag_rules_nixpkgs/nixpkgs/nixpkgs.bzl", line 149
_execute_or_fail(repository_ctx, <3 more arguments>)
File "/private/var/tmp/_bazel_runner/17d2b3954f1c6dcf5414d5453467df9a/external/io_tweag_rules_nixpkgs/nixpkgs/nixpkgs.bzl", line 318, in _execute_or_fail
fail(<1 more arguments>)
Cannot build Nix attribute 'nodejs'.
Command: [/Users/runner/.nix-profile/bin/nix-build, /private/var/tmp/_bazel_runner/17d2b3954f1c6dcf5414d5453467df9a/external/node_nix/nix/bazel.nix, "-A", "nodejs", "--out-link", "bazel-support/nix-out-link", "-I", "nixpkgs=/private/var/tmp/_bazel_runner/17d2b3954f1c6dcf5414d5453467df9a/external/nixpkgs/nixpkgs"]
Return code: 1
Error output:
src/main/tools/process-tools.cc:173: "setitimer": Invalid argument
```
* Update rules_scala
* .proto has been removed, use [ProtoInfo] instead
See
https://docs.bazel.build/versions/1.1.0/be/protocol-buffer.html#proto_library
* python3_nix add nix_file attribute
To avoid the following error
```
ERROR: /home/aj/tweag.io/da/da-bazel-1.1/BUILD:66:1: //:nix_python3_runtime depends on @python3_nix//:bin/python in repository @python3_nix which failed to fetch. no such package '@python3_nix//': Traceback (most recent call last):
File "/home/aj/.cache/bazel/_bazel_aj/5f825ad28f8e070f999ba37395e46ee5/external/io_tweag_rules_nixpkgs/nixpkgs/nixpkgs.bzl", line 149
_execute_or_fail(repository_ctx, <3 more arguments>)
File "/home/aj/.cache/bazel/_bazel_aj/5f825ad28f8e070f999ba37395e46ee5/external/io_tweag_rules_nixpkgs/nixpkgs/nixpkgs.bzl", line 318, in _execute_or_fail
fail(<1 more arguments>)
Cannot build Nix attribute 'python3'.
Command: [/home/aj/.nix-profile/bin/nix-build, "-E", "import <nixpkgs> { config = {}; overlays = []; }", "-A", "python3", "--out-link", "bazel-support/nix-out-link", "-I", "nixpkgs=/home/aj/.cache/bazel/_bazel_aj/5f825ad28f8e070f999ba37395e46ee5/external/nixpkgs/nixpkgs"]
Return code: 1
Error output:
error: anonymous function at /home/aj/.cache/bazel/_bazel_aj/5f825ad28f8e070f999ba37395e46ee5/external/nixpkgs/nixpkgs.nix:3:1 called with unexpected argument 'config', at (string):1:1
```
* rules_haskell unnamed string.split(_, maxsplit = _)
The keyword argument may no longer be named.
* string.replace(_, _, maxsplit = _) may not be named
* Move proto sources from deps to data
Fixes
```
ERROR: /home/aj/tweag.io/da/da-bazel-1.1/daml-lf/archive/BUILD.bazel:150:1: in deps attribute of scala_test rule //daml-lf/archive:daml_lf_archive_reader_tests_test_suite_src_test_scala_com_digitalasset_daml_lf_archive_DecodeV1Spec.scala: '//daml-lf/archive:daml_lf_1.6_archive_proto_srcs' does not have mandatory providers: 'JavaInfo'. Since this rule was created by the macro 'da_scala_test_suite', the error might have been caused by the macro implementation
```
* Define sha256 for haskell_ghc__paths
Bazel 1.1.0 fails on missing hashes.
* Disable --incompatible_windows_native_test_wrapper
* //compiler/daml-extension don't modify sources
Modifying sources in-place can cause issues on Windows, where build
actions are not sandboxed and changes on sources can affect other build
steps.
* bazel-genfiles --> bazel-bin
The bazel-genfiles symlink has been removed since Bazel 1.0.
See https://github.com/bazelbuild/bazel/issues/8651
* Mark dev_env_tool repository rule as configure
See
https://docs.bazel.build/versions/1.1.0/skylark/lib/globals.html#repository_rule
* Move data deps into data attribute
* Mark dev_env_tool as local = True
* Manually fetch @makensis_dev_env
* Update bazel-common to fix javadoc issues
Specifically, to fix the following error
```
ERROR: /home/aj/tweag.io/da/da-bazel-1.1/ledger-api/rs-grpc-bridge/BUILD.bazel:7:1: in javadoc_library rule //ledger-api/rs-grpc-bridge:rs-grpc-bridge_javadoc:
Traceback (most recent call last):
File "/home/aj/tweag.io/da/da-bazel-1.1/ledger-api/rs-grpc-bridge/BUILD.bazel", line 7
javadoc_library(name = 'rs-grpc-bridge_javadoc')
File "/home/aj/.cache/bazel/_bazel_aj/5f825ad28f8e070f999ba37395e46ee5/external/com_github_google_bazel_common/tools/javadoc/javadoc.bzl", line 27, in _javadoc_library
dep.java.transitive_deps
object of type 'JavaSkylarkApiProvider' has no field 'transitive_deps'
```
* Define Maven deps using rules_jvm_external
* Pin artifacts
* Remove bazel-deps generated targets
* Remove bazel-deps
* Switch to rules_jvm_external targets
* update bazel documentation
* pom_file: There are no more bazel-deps targets
* BAZEL-JVM.md `maven_install` typo
We currently use a custom cabal file for ghc-lib that has libffi in
the extra-libraries section so Hazel adds the headers. Forcing GHC to
use the bundled libffi should hopefully remove the need for this hack
which simplifies things.
* Refactor gRPC request cancellation support in gRPC-haskell
Previously, we had a separate constructor that took an additional
callback that had access to the client call so that we could use that
for cancellation. This PR removes the separate constructor and instead
changes the callback accepted by ClientReaderRequest (and for
consistency also the one accepted by ClientBiDiRequest) to pass the
client call.
This seems like a simpler solution so I’m more hopeful that we will be
able to upstream this change (this is the only API breaking change we
made afaik). There are two caveats here:
1. This will break existing consumers that use ClientReaderRequest. At
this stage in gRPC-haskell this is very reasonable and upstream
doesn’t seem to try very hard to avoid those.
2. The callback is called slightly later than the custom callback we
added before so you have to wait a bit longer until you can cancel
things. At least on our test suite that doesn’t seem to make a
difference.
* Fix async exception handling in grpc-haskell
This replaces the call to mask_ that I added in clientRequest as a
rudamentary fix by a proper fix that fixes the way async exceptions
are handled and should hopefully be more reasonable to upstream:
gRPC-haskell had a lot of cases where they did something like
```
do x <- allocateResource
f x `finally` cleanup x
```
That breaks if you get an exception after allocating the resource but
before the exception handler is installed by finally. I replaced all
of those instances by `bracket` where possible and one explicit call
to `mask` where it is not possible.
There is a bit of boilerplate here since the resource allocation
returns an `Either` but given that there doesn’t seem to be an
existing utils module where I could put this and I don’t want to
depend on `extra` for things like `whenRight` (since I want to
upstream this), I’ve just kept the boilerplate for now.
* Go back to building grpc-haskell-core using c2hs
This should hopefully avoid issues like the CSize vs CULong issue we
had a while back and might fix some of the issues we have been seeing
on CI.
I’m marking the Haskell ledger bindings as non-flaky for now so we can
see if the issues reappear.
* Fix path
* Fix c2hs runfiles
* s/basedir/dirname/
* Fix varname
* Remove fixme \o/
* Mark hs ledger bindings flaky again
For now, this only works on Linux (that’s a GHC limitation as far as I
know) and you have to enable it by setting the GHC_DWARF env var to a
non-empty string.
* Supporting producing sdist tarballs for the HS ledger bindings
The README.md has an explanation for how you can use this.
This should hopefully allow others to experiment with the bindings.
The gRPC library does not handle asynchronous exceptions properly and
ends up leaking things which causes various issues (segfaults,
assertion failures, weird backup poller timer messages, …) when
shutting down gRPC.
The switch to grpcShutdownBlocking is somewhat unrelated (the issues
happen with and without that and the fix seems to work both times) but
that function seems to be the right way to shut down gRPC in newer
versions and is what all the official language bindings switched to,
so I also made the switch.
I haven’t yet looked into the issues in the HS ledger bindings so not
sure if this helps with those as well.
* Move files in daml-foundations/daml-ghc to compiler/damlc
There is also a bit of refactoring going on to actually split things
apart into sensible targets. What is still missing is a cleanup of the
module hierarchy and a cleanup of the test targets but I’ll leave
those for separate PRs.
As a nice bonus, this also reduces dependencies between targets so it
will speed up compiles.
* Update .hie-bios
* Bazel: 0.24.0 -> 0.27.0
* Update rules_haskell for Bazel 0.27 compatibility
* Update bazel-deps and bazel-watcher
* Windows escape JVM flags
* load commands at top of .bzl file
Bazel 0.27 no longer allows load commands that are not at the beginning
of the file.
* Update Bazel rules
* subpackage boundary
* native is not defined in BUILD files
* yarn: @bazel/hide-bazel-files
Seems to be required since latest rules_nodejs version. Otherwise, yarn
fails with errors about existing BUILD or BUILD.bazel files.
* grpc-java plugin visibility
* Update fat_cc_library
* Nix Python3 toolchain
* Iteration over depset
* dev_env_package: Create symlinks one level deeper
To prevent symlinking the BUILD file as well. The nested BUILD file
confuses Bazel as of 0.27 and rules_nodejs cannot find the node
executable anymore.
* Update rules_nodejs
* Add managed_directories for node_modules
* hie-bios: Extract bazel-genfiles from bazel info
Bazel 0.27 changed the genfiles location which breaks the hie-core test
on macOS.
* update cc_wrapper to Bazel 0.27
* bazel info -> bazel info bazel-genfiles
* Fix typo in BUILD
Co-Authored-By: Stefano Baghino <43749967+stefanobaghino-da@users.noreply.github.com>
* Update rules_haskell and static GHC
Remove patches that have been upstreamed or are no longer required.
Update still required patches to match the new rules_haskell version.
Previously we patched rules_haskell to coerce GHC into using static
Haskell libraries in most places. In particular we moved hs-libraries
entries into extra-libraries entries in the package configuration files.
A much cleaner approach is to compile GHC with a static RTS, then GHC
will by itself choose to load static Haskell libraries.
* Remove haskell_cc_import
* da-hs-daml-cli -> daml-cli
* da-hs-damlc-app -> damlc-app
* windows: hanging GRPC FFI call problem resolution
* windows: fix visual test
* windows: enabled back again some of disabled daml-assitant tests
* marking daml_ghc_integration_test as large
* Fixes#1204: Release bindings and codegens to Maven Central.
Upload the Java and Scala Bindings with the respective code
generator binaries to Sonatype Open Source Repository
Host for synchronization with Maven Central.
* use with syntax in Daml
* attend to code review comments
* move ledger services implementation to own sub-module
* use ledger reset service when running tests for haskell-ledger-bindings
* expose LL.ClientCall in high level interface
* cancel streaming gRPC calls when attached stream is closed
* fix modification to gRPC-haskell so existing interface in preserved
* ClientCall and clientCallcancel were already available on HighLevel interface
There is no simple way to configure GCS to serve the desired security
headers, so instead the script will keep updating the existing s3
bucket.
Consequent changes:
- Add aws cli tool to dev-env
- Remove docs bucket from Terraform
* Update rules_haskell
- rules_haskell now handles the global package db within Bazel
https://github.com/tweag/rules_haskell/pull/859
- We no longer use the Nix provided c2hs. So, we drop it.
- Rename `ghcWithC2hs` to `ghcStatic` to clarify that that's where the
static linking patches are applied.
- Extend package-db patches to align Nix store paths with the new $out.
This works around a restriction in current rules_haskell, where
the paths in the package config files must have the same prefix as
the path to the package config files themselves.
- Don't exclude haskell libraries from extra-libraries entries.
* Drop redundant unix-compat override
This is a left-over from when the package was patched.
* Windows GHC bindist includes ffi header
* Drop unused language-c Nix override
This reverts commit 3d8acde916.
For some reason that commit seems to have resulted in a lot of
"unexpected end of file" errors during cache downloads. I do not know
what is going on here or how to fix it so let’s revert it for now.
* Move to using proto3-wire from upstream
* Move to upstream proto3-suite, with some custom patches in my fork
* Delete the BUILD.bazel for hte proto3 stuff, not used and the test was failing
* Delete the old proto3-wire and proto3-suite forks
* Delete proto3-wire
* Prettify BUILD.bazel files, sort the deps
* Remove some special cases from the license checker
* Delete unused Nix files from grpc-haskell
* Switch to upstream proto3-suite
* Make old-time work on Windows
* Formatting
* Patch rules_haskell to use a response file for -optP to avoid overflowing argument size limits on Windows
* Update 3rdparty/haskell/BUILD.old-time
Co-Authored-By: neil-da <35463327+neil-da@users.noreply.github.com>
* Update the comments in old-time
* Use the revised location of proto3-suite
* Inline c2hs expansion.
* Patch unix-compat for Windows
unix-compat fails on Windows due to missing version macros in hsc files.
This patches unix-compat inlining the effect of the corresponding
version macro evaluation.
* Add grpc-haskell to Windows CI
* Fix formatting
* Move unix-compat.patch
Moved to bazel_tools, where all other Bazel patches reside.
* Remove .chi files
Those don't need to be checked in.
* Add FIXME on checked in c2hs files.
This adds `ci/azure-cleanup`, containing a script that talks to azure pipelines, removing agents older than 25 hours in a specific pool.
Machines are meant to be killed after 24 hours anyway, make sure they're properly unregistered from Azure Pipelines, too.
By doing this, we don't need to unregister nodes manually on shutdown.
Idea is to execute this every time a new agent is provisioned, it has cloned the repo. We intend to clone the repo and pre-warm the caches there anyhow.
WIP until the repo fetching and cache pre-warming is present, too.
cc @zimbatm
### Pull Request Checklist
- [x] Read and understand the [contribution guidelines](https://github.com/digital-asset/daml/blob/master/CONTRIBUTING.md)
- [x] Include appropriate tests
- [x] Set a descriptive title and thorough description
- [x] Add a reference to the [issue this PR will solve](https://github.com/digital-asset/daml/issues), if appropriate
- [x] Add a line to the [release notes](https://github.com/digital-asset/daml/blob/master/docs/source/support/release-notes.rst), if appropriate
NOTE: CI is not automatically run on non-members pull-requests for security
reasons. The reviewer will have to comment with `/AzurePipelines run` to
trigger the build.
* Add buildifier targets.
The tool allows to check and format BUILD files in the repo.
To check if files are well formatted, run:
bazel run //:buildifier
To fix badly-formatted files run:
bazel run //:buildifier-fix
* Cleanup dade-copyright-headers formatting.
* Fix dade-copyright-headers on files with just the copyright.
* Run buildifier automatically on CI via 'fmt.sh'.
* Reformat all BUILD files with buildifier.
Excludes autogenerated Bazel files.
* nix: add the more providers to terraform
* docs: make tarballs more reproducible
* ci: use the linux-pool pool
* ci: tweak the nix installation
handle the case where the user is root and on ubuntu
* infra: terraform fmt
* infra: add Azure Pipeline agents
* ci: only enable linux-pool for internal PRs
* Add filepath as a dependency of proto3-suite
* Make path splitting work on Windows
* Add directory as a dependency
* Avoid generating new files each time with proto-buf
* Add the libraries to Bazel too
* Fetch status.proto from remote, simplify JS gRPC codegen
Fetch the `status.proto` file (part of the standard gRPC distribution)
from a distribution channel. _Moreover_, use the recently introduced
`proto_gen` rule to simplify how the gRPC code for the Node.js bindings
are generated (and remove the need to have `google/rpc/status.proto`
locally in the repository.
* Add plugin_runfiles option to proto_gen
This allows use to add additional files to the bazel sandbox so that
plugins can refer to them. This will subsequently be used by the
protoc-gen-doc plugin.
Also, pass the plugin options via --name_opt parameter.
* Add missing status.proto dependency /language-support/java and /ledger
* Build proto docs using the proto_gen rule
To make this work, I had to turn on the bazel build flag
`--protocopt=--include_source_info` because we cannot turn enable this
flag only for specific build rules.
* Make /ledger-api/grpc-definitions:docs public again
* Revert to the old style of passing plugin arguments to --name_out=options:path
* Suppress output of unzipping
* Fix link for google.rpc.Status in proto-docs
