[vcpkg-scripts] Update to CMake 4.0.0-rc4#44273
[vcpkg-scripts] Update to CMake 4.0.0-rc4#44273Neumann-A wants to merge 25 commits intomicrosoft:masterfrom
Conversation
|
@BillyONeal Is it intended that CI is not printing the actual download failure due to the hash mismatch (since the hash has only be updated for windows & amd64)? |
I'm not sure I understand the question. It won't report failures to download things it never attempts to download. |
I updated the version to 4.0.0-rc3 so shouldn't it always attempt to download that version? However, it just gives an error that it cannot find cmake where it expects it to find it. I expected it to print a hash mismatch error instead. Ah maybe that is the same issue as everywhere that it does not download it if there is a hash match because i didn't set it to 0. |
|
It first tries to download the thing through the asset cache via the given sha, if this fails it tries to download from the original url. |
|
@autoantwort I tend to forget that fact sometimes. However, can you take a look at the osx failure. I assume something is wrong with the download path for some reason? (The SHA is now changed so it should download the new version) |
|
Remaining: mac really doesnt like cmake 4 |
|
@dg0yt: |
|
Did you notice that android is now dynamic crt, API Level 28? |
|
PR is basically green and thus theoretically ready for review. |
|
@ras0219-msft @vicroms @BillyONeal @JavierMatosD discussed this today. I'm sorry this comment doesn't necessarily give you a path forward here @Neumann-A ; I thought it important to record the thought process even if we don't really have a good answer right now. We have a problem in that there doesn't seem to be a great answer to provide the kind of compat guarantees we would normally want to provide given that CMake has done this to us. Originally, we were going to say something like:
but after testing that Potential ideas. This is a brainstorm, not picking one:
We probably won't have landed on a path forward before 4.x is actually out. We're curious what the Linuxes do here; do they add |
|
In the meantime CMake 4 has been released.
I could override cmake_policy? And ignore policy settings which would error with the newer CMake versions. It wouldn't help however with the slight syntax errors projects seem to have. I could also be sneaky and first upgrade all ports and then hide stability behind only changing the cmake version ;)
Doesn't this mean to basically have two verisons of CMake? One running vcpkg and run running the ports? Also this would imply a tool port for CMake?
Most of the errors are on macosx so they probably do not care? And setting set(ENV{CMAKE_POLICY_VERSION_MINIMUM} 3.5) is basically all that is needed to make nearly all projects build. |
30c59ad ci: Drop confusing comment (Hennadii Stepanov) ef00a28 ci: Add workaround for vcpkg's libevent package (Hennadii Stepanov) Pull request description: This PR is necessary for Windows GHA [images](https://github.com/actions/runner-images/blob/win22/20250330.1/images/windows/Windows2022-Readme.md), which provide CMake >= 4.0. The idea has been taken from microsoft/vcpkg#44273 (comment). ACKs for top commit: maflcko: lgtm ACK 30c59ad pinheadmz: ACK 30c59ad Tree-SHA512: 898a616a40c1aaec69d8a329d56f887cacec71b9588842feef2b689fbd21a12d8f8b3b4053a6373332008668960b9df7c71a44fcd97e637e250705daf1215c7f
|
Would you be OK if we cherry pick out the broadly compatible port fixes in here (as in BillyONeal@d487095 ) into a separate PR, or would you be willing to do that? I don't want to take credit for your work but I also don't think we need to block those genuine improvements behind a call to force upgrade folks to CMake 4.x |
|
Merged the 'cherry pick' commit d6f1a7a @Neumann-A |
|
Most of the changes here were merged into #46911 . I made sure @Neumann-A was marked as a co-author when merging it. |
|
Placing in draft until file conflicts are resolved. |
Fixes #43816