Trying to fix sfinae problems by reverting vcpkg upgrade (Iroha 1)

[baziorek]

Description

Some time ago vcpkg package manager was bumped to newest version (commit: fdc2c1e). It caused dependencies to upgrade a lot so project stopped to compile. I was able to fix compilation of binaries (commit: 3b40f22) but I was unable to fix compilation of tests (SFINAE problem) after some time of trying.

So I decided that we need to make project compiling with not the most professional way, but it should work (and be first step for more professional way). The way is to revert those commits:

1. fdc2c1e
2. 3b40f22

After those reverts it compiled locally (I used docker: docker run -it --name='Ubuntu_under_test_22.04' -v ${PWD}/./:/home/ --workdir=/home/ ubuntu:22.04 with g++ (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0) - both binaries and tests compiled with commands:

cmake -B build -DCMAKE_TOOLCHAIN_FILE=$PWD/vcpkg-build/scripts/buildsystems/vcpkg.cmake . -DCMAKE_BUILD_TYPE=RELEASE   -GNinja -DUSE_BURROW=OFF -DUSE_URSA=OFF -DTESTING=ON -DPACKAGE_DEB=OFF  # testing ON
cmake --build ./build --target all  # target = all

For now I want to check if CI is compiling, that is why the PR is not ready for review yet.

Difference between versions

1. Old version of vcpkg:

$ ./vcpkg-build/vcpkg version
Vcpkg package management program version 2021-01-13-unknownhash
$ git log -1
commit e8dbfcf6797a270ed5be8550248f7fe4fe5dec79 (HEAD)
Author: Matthias C. M. Troffaes <[email protected]>
Date:   Tue May 4 07:14:55 2021 +0100

    [ffmpeg] soxr dependency fix (#17299)
    
    * [ffmpeg] soxr only makes sense with swresample, so add as dependency
    
    * [ffmpeg] bump port version
    
    * [ffmpeg] x-add-version

2. New version of vcpkg:

$ ./vcpkg-build_newest/vcpkg version
vcpkg package management program version 2023-02-16-12e657924d99511514c0287ca5ce46882d3657c7
$ git log -1
commit a7b6122f6b6504d16d96117336a0562693579933 (HEAD, tag: 2023.02.24)
Author: Frank <[email protected]>
Date:   Sat Feb 25 06:08:21 2023 +0800

    [fmilib] Change to the github and update to fix bug of libexpat (#29805)

Version compare in table (generated by ChatGpt, so there can be mistakes):

Library	Before Upgrade	After Upgrade
abseil	2021-03-24	20230125.0
benchmark	1.5.2	1.7.1
boost-accumulators	1.75.0	1.81.0#2
boost-algorithm	1.75.0	1.81.0#2
boost-align	1.75.0	1.81.0#2
boost-any	1.75.0	1.81.0#2
boost-array	1.75.0	1.81.0#2
boost-asio	1.75.0#1	1.81.0#2
boost-assert	1.75.0	1.81.0#2
iroha-ed25519	2.0.1	2.0.1
liblzma	5.2.5#2	5.4.1#1
libpq	12.2#16	14.4#3
libpq[openssl]	...	...
libpq[zlib]	...	...
lz4	...	1.9.4#1
nlohmann-json	3.9.1	3.11.2
openssl	1.1.1k	3.0.8
prometheus-cpp	0.12.2	1.1.0
prometheus-cpp[compression]	...	...
prometheus-cpp[pull]	...	...
protobuf	3.15.8	3.21.12
rapidjson	2020-09-14	2022-06-28#3
re2	2020-10-01	2023-02-01
rocksdb	6.14.6	7.9.2
rocksdb[zlib]	...	...
rxcpp	4.1.0-1	4.1.1#1
soci	4.0.1#3	4.0.3
soci[boost]	...	...
soci[postgresql]	...	...
spdlog	1.8.5#2	1.11.0
upb	2020-12-19#1	2022-06-21
upb[codegen]	...	...
vcpkg-cmake-config	2021-02-26#1	2022-02-06#1
vcpkg-cmake-get-vars	...	2022-12-16
vcpkg-cmake	2021-02-28#1	2022-12-22
zlib	1.2.11#10	1.2.13
zstd	1.4.9	1.5.4

vcpkg_libraries_new.txt
vcpkg_libraries_old.txt

Future plans:

If my fix works packages can be upgraded one by one until we face the SFINAE problem again - that package/packages would stay in older version. Of course this can be done in another PR.

Linked issue

Closes #{issue_number}

Benefits

Iroha 1 starts compilling again and another release 1.6 would be possible.

Checklist

• I've read CONTRIBUTING.md
• I've used the standard signed-off commit format (or will squash just before merging)
• All applicable CI checks pass (or I promised to make them pass later)
• (optional) I've written unit tests for the code changes
• I replied to all comments after code review, marking all implemented changes with thumbs up

[baziorek]

I checked the CTest problem:

I run the test locally and it passes:

./build$ ../test/tool/test_tool_iroha_wsv_diff.sh -block_store_path /home/iroha/build/test_data/block_store/ -bin_dir /home/iroha/build/bin/

Then I downloaded && checked logs from CTest and it looks that the test is passing:

2023-08-31T00:13:44.7806214Z [  PASSED  ] 3. test_odd_domain_account_role
2023-08-31T00:13:44.7806359Z ----------------------------------------------------
2023-08-31T00:13:44.7806456Z TOTAL 3 tests were executed.
2023-08-31T00:13:44.7806552Z PASSED ALL 3 of 3 tests: 
2023-08-31T00:13:44.7806650Z    [  PASSED  ] 1. test_equal_wsv
2023-08-31T00:13:44.7806785Z    [  PASSED  ] 2. test_wrong_permissions_and_asset_amount
2023-08-31T00:13:44.7806906Z    [  PASSED  ] 3. test_odd_domain_account_role
2023-08-31T00:13:44.7807001Z HOORAY. all tests PASSED.
2023-08-31T00:13:44.7807010Z 
2023-08-31T00:13:44.7807016Z 
2023-08-31T00:13:44.7807136Z 75% tests passed, 1 tests failed out of 4

So all 3 tests are passing, but somewhere test executor expects 4 tests.

EDIT: Interesting thing is that there are both tests:

1. "build-UD (ubuntu gcc-10 Debug normal)"
2. "build-UR (ubuntu gcc-10 Release normal)"
   Debug is passing, because it is not executing the test (log file is much shorter).
   So to fix the issue somebody should have devops knowledge. If nobody has time we can add tool_iroha_wsv_diff to iroha/.github/TESTS_ALLOWED_TO_FAIL?

[baziorek]

In the last commit (38d725b) I was trying to fix the test tool_iroha_wsv_diff. To fix this it was necessarily to change iroha_wsv_diff tool. The tool is comparing WSV between Postgres and RocksDB. The test failure was because it contained check:

assert(schema_version == "1#4#0" &&
                   "This version of iroha_wsv_diff can check WSV in RocksDB of version 1.4.0 only");

As I checked now: Iroha 1.5 (https://github.com/hyperledger/iroha/releases/tag/1.5.0) does not change anything in schema after version 1.4. So I've added extra parameter which disables checking of schema version. This argument is used in tests. It is not perfect solution. In future probably the tool needs to be updated to support scheme after changes.