Refine release process #1310

pull jonasnick wants to merge 4 commits into bitcoin-core:master from jonasnick:refine-release-process changing 1 files +33 −6
  1. jonasnick commented at 2:21 pm on May 12, 2023: contributor
    Fixes #1176
  2. in doc/release-process.md:33 in 74826e857b outdated 
    28+gcc -o ecdsa ecdsa.c -I $dir/include -L $dir/lib64/ -l secp256k1
29+```
30+3. Check installation with CMake:
31+```
32+dir=$(mktemp -d)
33+mkdir build && cd build && cmake -DCMAKE_INSTALL_PREFIX=$dir .. && make install && ls -l $dir/include $dir/lib64
    real-or-random commented at 2:28 pm on May 12, 2023: 
    0mkdir build && cmake -B build -DCMAKE_INSTALL_PREFIX=$dir && cmake --build build --target install && ls -l $dir/include $dir/lib*
  3. jonasnick force-pushed on May 12, 2023
  4. jonasnick marked this as ready for review on May 12, 2023
  5. in doc/release-process.md:28 in c0751c0115 outdated 
    23+```
24+2. Check installation with autotools:
25+```shell
26+dir=$(mktemp -d)
27+./configure --prefix=$dir && make install && ls -l $dir/include $dir/lib
28+gcc -o ecdsa examples/ecdsa.c -I $dir/include -L $dir/lib/ -l secp256k1 && ./ecdsa
    real-or-random commented at 3:36 pm on May 12, 2023: 
    0gcc -o ecdsa examples/ecdsa.c -I $dir/include -L $dir/lib/ -static -l secp256k1 && ./ecdsa
    real-or-random commented at 3:41 pm on May 12, 2023: 
    0gcc -o ecdsa examples/ecdsa.c -I $dir/include -L $dir/lib/ -static -l secp256k1 && ./ecdsa
    hebasto commented at 10:21 pm on May 12, 2023:

    We could test the libsecp256k1.pc file as well:

    0gcc -o ecdsa examples/ecdsa.c $(PKG_CONFIG_PATH=$dir/lib/pkgconfig pkg-config --cflags --libs libsecp256k1) && ./ecdsa
  6. in doc/release-process.md:33 in c0751c0115 outdated 
    28+gcc -o ecdsa examples/ecdsa.c -I $dir/include -L $dir/lib/ -l secp256k1 && ./ecdsa
29+```
30+3. Check installation with CMake:
31+```shell
32+dir=$(mktemp -d)
33+mkdir build && cmake -B build -DCMAKE_INSTALL_PREFIX=$dir && cmake --build build --target install && ls -l $dir/include $dir/lib*
    real-or-random commented at 3:41 pm on May 12, 2023: 
    0mkdir build && cmake -B build -DCMAKE_INSTALL_PREFIX=$dir -DBUILD_SHARED_LIBS=OFF && cmake --build build --target install && ls -l $dir/include $dir/lib*
    hebasto commented at 9:03 am on May 15, 2023:
    cmake -B build ... creates the build directory if needed. So, it seem the only purpose of mkdir build is to prevent building in a non-clear binary tree. Perhaps we can use rm -rf build for that purpose (which I use in my everyday CMake workflow)?
    jonasnick commented at 2:15 pm on May 23, 2023:
    I don’t want to rm -rf the user’s files.
    jonasnick commented at 2:17 pm on May 23, 2023:
    @real-or-random why wouldn’t we want to build the shared library?
    real-or-random commented at 3:51 pm on May 23, 2023:

    Since #1230, our CMake builds are either shared or static (but not both), because that’s more common and simpler in CMake. And variable BUILD_SHARED_LIBS is the CMake convention for this. So -DBUILD_SHARED_LIBS=OFF doesn’t only enable static libs, it enables static builds. But

    But yeah, the rpath thing does the job as well and is probably a bit cleaner.

    real-or-random commented at 1:09 pm on May 24, 2023:

    I don’t want to rm -rf the user’s files.

    nit: If you want to touch this again, what about this variant for additional convenience?

    0build=$(mktemp -d)
1cmake -B $build -DCMAKE_INSTALL_PREFIX=$dir && cmake --build $build --target install && ls -l $dir/include $dir/lib*```
  7. in doc/release-process.md:32 in eb057be9d5 outdated 
    27+```
28+3. Check installation with CMake:
29+```shell
30+dir=$(mktemp -d)
31+mkdir build && cmake -B build -DCMAKE_INSTALL_PREFIX=$dir && cmake --build build --target install && ls -l $dir/include $dir/lib*
32+gcc -o ecdsa examples/ecdsa.c -I $dir/include -L $dir/lib*/ -l secp256k1 && ./ecdsa
    hebasto commented at 9:59 pm on May 12, 2023:
    The last command, ./ecdsa, needs the LD_LIBRARY_PATH variable to specify the path to the libsecp256k1.so.2 library.
    real-or-random commented at 8:24 am on May 13, 2023:

    Right, that’s the alternative to static linking I suggested above.

    Maybe all of this sanity-checking should go to a small script that then can also runs on CI then?

    hebasto commented at 9:39 am on May 13, 2023:
    #1310 (review)
  8. in doc/release-process.md:27 in c0751c0115 outdated 
    22+make distcheck
23+```
24+2. Check installation with autotools:
25+```shell
26+dir=$(mktemp -d)
27+./configure --prefix=$dir && make install && ls -l $dir/include $dir/lib
    hebasto commented at 10:06 pm on May 12, 2023: 
    0./autogen.sh && ./configure --prefix=$dir && make install && ls -l $dir/include $dir/lib
  9. hebasto cross-referenced this on May 12, 2023 from issue release: Prepare for 0.3.2 by real-or-random
  10. in doc/release-process.md:22 in c0751c0115 outdated 
    17+## Sanity Checks
18+Perform these checks before creating a release:
19+
20+1. Ensure `make distcheck` doesn't fail.
21+```shell
22+make distcheck
    hebasto commented at 10:47 pm on May 12, 2023: 
    0./autogen.sh && ./configure && make distcheck
  11. in doc/release-process.md:42 in c0751c0115 outdated 
    38 
39 1. Open a PR to the master branch with a commit (using message `"release: prepare for $MAJOR.$MINOR.$PATCH"`, for example) that
40-   * finalizes the release notes in [CHANGELOG.md](../CHANGELOG.md) (make sure to include an entry for `### ABI Compatibility`),
41+   * finalizes the release notes in [CHANGELOG.md](../CHANGELOG.md) by
42+       * adding a section for the release (make sure that the version number is a link to a diff between the previous and new version) and
43+       * including an entry for `### ABI Compatibility` if it doesn't exist,
    hebasto commented at 10:54 pm on May 12, 2023:
    Speaking of ABI compatibility, does it make sense to specify SONAME on Linux, for example libsecp256k1.so.2?
    real-or-random commented at 8:30 am on May 13, 2023:

    You mean, saying that the shared library is libsecp256k1.so.2? I like this idea. It makes the ABI version very explicit.

    Then we probably want to say bikeshed a bit about the exact phrasing; Does this apply to macOS, too? Do we want to say something for Windows? Etc… (So that should not hold up 0.3.2 now, I think.)

    hebasto commented at 9:27 am on May 13, 2023:

    You mean, saying that the shared library is libsecp256k1.so.2?

    To be pedantic, the shared library’s soname is libsecp256k1.so.2.

    I like this idea. It makes the ABI version very explicit.

    That is my point :)

    Then we probably want to say bikeshed a bit about the exact phrasing; Does this apply to macOS, too? Do we want to say something for Windows? Etc… (So that should not hold up 0.3.2 now, I think.)

    I’d skip them for now. While an ABI version is a part of DYLIB and DLL names, AFAIK, linking processes on macOS and Windows do not rely on it.

  12. in doc/release-process.md:34 in c0751c0115 outdated 
    29+```
30+3. Check installation with CMake:
31+```shell
32+dir=$(mktemp -d)
33+mkdir build && cmake -B build -DCMAKE_INSTALL_PREFIX=$dir && cmake --build build --target install && ls -l $dir/include $dir/lib*
34+gcc -o ecdsa examples/ecdsa.c -I $dir/include -L $dir/lib*/ -l secp256k1 && ./ecdsa
    hebasto commented at 9:38 am on May 13, 2023:

    Using -rpath linker options helps as well:

    0gcc -o ecdsa examples/ecdsa.c -I $dir/include -L $dir/lib*/ -l secp256k1 -Wl,-rpath,"$ORIGIN/$dir/lib" && ./ecdsa

    Tbh, I like this approach the most (in comparison to others).

    jonasnick commented at 2:16 pm on May 23, 2023:
    I added this variant, but without $ORIGIN because $dir already specifies a full path.
  13. real-or-random commented at 6:10 pm on May 13, 2023: contributor
    We could include the following: remove the [Unreleased] section header in Changelog in the preparation PR, and readd it in the PR after the release. We did this for 0.3.2, and I think it makes sense.
  14. real-or-random added this to the milestone 0.3.3 (or 0.4.0) on May 13, 2023
  15. jonasnick force-pushed on May 23, 2023
  16. jonasnick commented at 2:18 pm on May 23, 2023: contributor
    Updated the branch. Thanks @hebasto @real-or-random for the suggestions.
  17. in doc/release-process.md:22 in 699c812767 outdated 
    17+## Sanity Checks
18+Perform these checks before creating a release:
19+
20+1. Ensure `make distcheck` doesn't fail.
21+```shell
22+./autogen.sh && ./configure && make distcheck
    real-or-random commented at 3:58 pm on May 23, 2023: 
    0./autogen.sh && ./configure --enable-dev-mode && make distcheck

    This would enable all modules, which would make the distcheck cover them. If you are concerned that dev mode is not the right thing here, we could also check once with and without --enable-dev-mode.

  18. in doc/release-process.md:34 in 699c812767 outdated 
    29+```
30+3. Check installation with CMake:
31+```shell
32+dir=$(mktemp -d)
33+mkdir build && cmake -B build -DCMAKE_INSTALL_PREFIX=$dir -DBUILD_SHARED_LIBS=OFF && cmake --build build --target install && ls -l $dir/include $dir/lib*
34+gcc -o ecdsa examples/ecdsa.c -I $dir/include -L $dir/lib*/ -l secp256k1 -Wl,-rpath,"$dir/lib" && ./ecdsa
    real-or-random commented at 4:00 pm on May 23, 2023:

    Okay, I think -DBUILD_SHARED_LIBS=OFF (= static library) together with -rpath (= search path for dynamic linker) does not make sense.

    I think we should just stick either to the shared library for both build systems, or, if we want some more variety in the sanity checks, use the static library here. (And drop the rpath then.)

    jonasnick commented at 4:15 pm on May 23, 2023:
    The current PR uses the shared library for both build systems.
    real-or-random commented at 4:34 pm on May 23, 2023:

    The current PR uses the shared library for both build systems.

    Are you sure? I don’t think so. These CMake commands here shouldn’t build a shared library.

    jonasnick commented at 4:51 pm on May 23, 2023:
    Oh sorry, I thought I had removed the -DBUILD_SHARED_LIBS=OFF after testing it.
    jonasnick commented at 4:52 pm on May 23, 2023:
    Fixed now
  19. real-or-random commented at 4:01 pm on May 23, 2023: contributor

    ACK mod nits.

    I still need to test the sanity check instructions. I will do once the nits have been addressed.

  20. jonasnick force-pushed on May 23, 2023
  21. in doc/release-process.md:34 in fcb2b6e2ef outdated 
    29+```
30+3. Check installation with CMake:
31+```shell
32+dir=$(mktemp -d)
33+mkdir build && cmake -B build -DCMAKE_INSTALL_PREFIX=$dir && cmake --build build --target install && ls -l $dir/include $dir/lib*
34+gcc -o ecdsa examples/ecdsa.c -I $dir/include -L $dir/lib*/ -l secp256k1 -Wl,-rpath,"$dir/lib" && ./ecdsa
    hebasto commented at 8:53 am on May 24, 2023:
    Add $dir/lib64 to -rpath as well?
  22. hebasto approved
  23. hebasto commented at 8:53 am on May 24, 2023: member
    ACK fcb2b6e2effffc8d9898de44bee72162c57aaf71
  24. jonasnick force-pushed on May 24, 2023
  25. hebasto approved
  26. hebasto commented at 12:39 pm on May 24, 2023: member
    re-ACK aeb6fc40cee2f21b15ff69acf0d2adb63b9a3b6d, tested on Ubuntu 22.04 and Fedora 38.
  27. in doc/release-process.md:27 in aeb6fc40ce outdated 
    22+./autogen.sh && ./configure --enable-dev-mode && make distcheck
23+```
24+2. Check installation with autotools:
25+```shell
26+dir=$(mktemp -d)
27+./autogen.sh && ./configure --prefix=$dir && make install && ls -l $dir/include $dir/lib
    real-or-random commented at 1:00 pm on May 24, 2023:

    I need this when I run step 2 immediately following step 1:

    0./autogen.sh && ./configure --prefix=$dir && make clean && make install && ls -l $dir/include $dir/lib

    Everything else works fine for me.

    hebasto commented at 1:26 pm on May 24, 2023:
    Maybe use git clean -xdff to start from the clean repo?
    real-or-random commented at 1:29 pm on May 24, 2023:

    Maybe use git clean -xdff to start from the clean repo?

    I think this will potentially kill some files that people still need (the same reason why @jonasnick didn’t like rm -rf build. We could start in a fresh clone or worktree, but that may be overkill.

    jonasnick commented at 1:42 pm on May 24, 2023:
    I hope that this will solve itself once it runs as part of CI.
    jonasnick commented at 1:43 pm on May 24, 2023:
    Added make clean
  28. real-or-random commented at 1:09 pm on May 24, 2023: contributor
    ACK mod nits
  29. release process: add sanity checks 165206789b
  30. release process: mention targeted release schedule 79fa50b082
  31. release process: fix process for maintenance release 6348bc7eee
  32. release process: clarify change log updates ad84603297
  33. jonasnick force-pushed on May 24, 2023
  34. real-or-random approved
  35. real-or-random commented at 1:46 pm on May 24, 2023: contributor

    ACK ad846032973cc1afd360613626c4e475bba66f56

    (We could rerun CI here to make it green but whatever, this PR touches only an .md file…)

  36. hebasto approved
  37. hebasto commented at 1:49 pm on May 24, 2023: member
    re-ACK ad846032973cc1afd360613626c4e475bba66f56
  38. real-or-random merged this on May 24, 2023
  39. real-or-random closed this on May 24, 2023
  40. vmta referenced this in commit e1120c94a1 on Jun 4, 2023
  41. sipa referenced this in commit 901336eee7 on Jun 21, 2023
  42. vmta referenced this in commit 8f03457eed on Jul 1, 2023
  43. hebasto referenced this in commit 270d2b37b8 on Jul 21, 2023
  44. jonasnick cross-referenced this on Jul 26, 2023 from issue Upstream PRs 1314, 1317, 1318, 1316, 1327, 1310, 1328, 1333, 1330, 1334, 1337, 1341, 1339, 1350, 1349, 1338, 1129, 1347, 1336, 1295, 1354, 1355, 1356 by jonasnick


jonasnick real-or-random hebasto

Milestone
0.3.3 (or 0.4.0)

github-metadata-mirror

This is a metadata mirror of the GitHub repository bitcoin-core/secp256k1. This site is not affiliated with GitHub. Content is generated from a GitHub metadata backup.
generated: 2024-10-30 05:15 UTC
This site is hosted by @0xB10C
More mirrored repositories can be found on mirror.b10c.me