[mpc/mpfr] Add new port / update mpfr#13081
Conversation
|
@BillyONeal: depends VM update: 2c2f1e4 . AX_PTHREAD missing. https://www.gnu.org/software/autoconf-archive/ax_pthread.html |
Co-authored-by: NancyLi1013 <46708020+NancyLi1013@users.noreply.github.com>
|
mipr:x64-linux: |
|
@NancyLi1013: Why requires author-response? autoconf-archive is missing which is why the error is there. See comment #13081 (comment) |
|
Thanks for your reply. @Neumann-A |
|
/azp run |
|
Azure Pipelines successfully started running 1 pipeline(s). |
|
The failure on x86-windows: |
|
@NancyLi1013: Probably due to the changes to msys. Will only touch it if linux vm has the required deps installed otherwise CI will just be red any way. |
|
/azp run |
|
No pipelines are associated with this pull request. |
|
@Neumann-A I will try to repro it manually. icu doesn't have any dependencies. Totally there are 248 ports were built before building icu. |
|
Although I installed these all 248 ports before install icu, I still can't repro this regression. |
|
Install scintilla before install icu, this regression can't repro too. @BillyONeal Do you have any suggestions to repro this? |
|
log about icu regression: |
|
Yeah green ... but i am curious what happened to the regression |
|
@Neumann-A Anyway, I recorded that in this PR. |
| { | ||
| "name": "gmp", | ||
| "version-string": "6.2.1-2", | ||
| "version-string": "6.2.1", |
There was a problem hiding this comment.
Since this already uses a manifest file, why not use version-semver or version versioning schemes?
There was a problem hiding this comment.
Why should it? This is a vcpkg internal update.
As such port-version is the correct thing and should have been from the beginning.
There was a problem hiding this comment.
In my opinion we should start switching packages to their correct versioning schemes.
Since "version-string" are non-orderable (except by their port version number) by design, continuing to use them will seriously limit the usefulness of "version>=" constraints.
There was a problem hiding this comment.
I am just doing what vcpkg x-add-version does. If you want "version": "6.2.1", here please tell/teach the command to do that.
full log Will try to rerun the test. |
|
It seems a baseline regression. |
|
Already merged from master. |
[mpc/mpfr] Add new port / update mpfr (microsoft#13081)
|
@BillyONeal this is an example of what I was trying to discuss with you earlier: this PR changed a dependency, from mpir to gmp, without any additional note, and broke downstream check in CI for other ports without anyone noticing. The problem is that now, due to already accepted failures in baseline, all ports that were depending on mpir and mpfr are now broken by default, since gmp cannot be installed if mpir is installed. TL;DR : try to install openmvs now. It depends on both gmp and mpir and so it's bound to fail. It was not before. It is completely untested by ci now, and it was not before. This is only an example. |
|
That's similar to what I experience always with the opencv port. Somehow, externally it gets broken and no-one gets notified about that. That just means more work to fix parallel unrelated ports every time I want to update it (and make it really work of course, not just pretending it) |
probably due to #13081 (comment). Need to install Trying to fix it in #15860 but I really need an inverse dependency lookup for those issues (e.g. |
|
Yes, you should put a note in the port file for linux and osx to install autoconf-archive (both on apt and in brew). |
might depend on #12936 if windows CI fails