Re: help needed: Windows CI broken

On 2026-06-02 10:45 a.m., Paul Lodge wrote:
> Hi All,
> 
> It might be in the best interest to move forward with Vlads suggestion 
> for now and update the CMAKE_MINIMUM_REQUIRED but I'd leave the 'i' 
> variable unset, as it's not an optimal solution it's ok as a quick fix 
> but could potentially cause headaches in the future.
> 
> I think that the Windows side of things in general (CMake, tests etc) 
> need a little TLC, I don't mind looking into and taking on the task.
> 
> Best
> 
> Paul

The CMAKE_MINIMUM_REQUIRED and the unset(i) are required to compile 
branch 2.4.x with MSVC (I do this manually, nothing to do with 
Docker/GitHub runner images etc).  I think there is a mixup in 
discussions here, as initially Joe was pointing to a branch 2.4.x run 
that was failing (and gave an error regarding to that "i"), and I 
provided the workaround for that branch, with the 
CMAKE_MINIMUM_REQUIRED, and unset(i).

Trunk is a different story (as I mentioned earlier, there are missing 
file errors/compile errors when compiling on Windows with MSVC now).

-jeff




> 
> 
> On 02/06/2026 09:19, Vladimir Chlup wrote:
>> Hello,
>>
>> The issue is most likely caused by the change of the default GH 
>> Actions Windows image. See https://github.com/actions/runner-images/ 
>> issues/14017 You can try using `windows-2022` instead of the current 
>> `windows-latest`. It should behave as before.
>>
>> Alternatively, I have prepared a PR with a fix here: https:// 
>> github.com/apache/httpd/pull/659
>>
>> Regards,
>> Vladimír Chlup
>>
>>
>> On Mon, Jun 1, 2026 at 6:45 PM Jeff McKenna 
>> <[email protected]> wrote:
>>
>>     On 2026-06-01 12:53 p.m., Jeff McKenna wrote:
>>     > On 2026-06-01 12:39 p.m., Joe Orton wrote:
>>     >> On Mon, Jun 01, 2026 at 12:24:14PM -0300, Jeff McKenna wrote:
>>     >>> On 2026-06-01 9:54 a.m., Joe Orton wrote:
>>     >>>> Windows builds have been failing in CI for a while, I think
>>     because of
>>     >>>> some changes in the GHA images.
>>     >>>>
>>     >>>> I burnt some Claude tokens trying to fix the trunk to no
>>     avail [1], but
>>     >>>> I have zero interest in this. It will be ... frustrating...
>>     if people
>>     >>>> start voting against future 2.4.x rcs because Windows builds
>>     are broken
>>     >>>> in some way which was trivial to catch in CI and fix ahead of
>>     time.
>>     >>>>
>>     >>>> Example trunk failure: https://github.com/apache/httpd/actions/
>>     >>>> runs/25912053842/job/76159381743
>>     >>>>
>>     >>>> 2.4.x failure is "simpler" because of the cmake version mismatch:
>>     >>>> https://github.com/apache/httpd/actions/runs/25488877019/
>>     >>>> job/74791446628
>>     >>>> but working around that as suggested did not help.
>>     >>>>
>>     >>>> Regards, Joe
>>     >>>>
>>     >>>> [1] https://github.com/notroj/httpd/tree/refs/heads/maybe-fix-
>>     >>>> windows-ci
>>     >>>
>>     >>> Branch 2.4.x needs this change to be backported: https://
>>     github.com/
>>     >>> apache/httpd/commit/08138ac986851217ff8f03b495d12ceca9051e24
>>     >>
>>     >> Does that definitely work? I tried -
>>     DCMAKE_POLICY_VERSION_MINIMUM=3.5
>>     >> and it failed:
>>     >> https://github.com/apache/httpd/actions/runs/26754488163/
>>     job/78850661003
>>     >
>>     > Ah!  I didn't notice that attempt.  To answer your question: yes
>>     that
>>     > CMake change that I made works.  The run that you pointed to you
>>     here
>>     > makes it past that (thanks to that fix) and now breaks on a
>>     different
>>     > issue.
>>     >
>>     > I had made a change locally to avoid that error: Unexpected
>>     enablement
>>     > value "i" for mod_authnz_ldap
>>     >
>>     > But I wasn't proud of my workaround, and now I realize that I
>>     should
>>     > create a pull request for my workaround for that "i" error (if I
>>     > remember, I had to "unset(i)" outside of a for loop, to avoid that
>>     > error.  Will file a new pull request now (to the 2.4.x branch,
>>     if you
>>     > already applied the other fix that I mentioned).
>>     >
>>     >>
>>     >>> PS. Suggest to use the Backport action in the httpd repo (we
>>     use this
>>     >>> and it
>>     >>> works beautifully https://github.com/MapServer/MapServer/blob/
>>     >>> main/.github/workflows/backport.yml
>>     >>> )
>>     >>
>>     >> Sounds useful.... with writeable git ;)
>>     >>
>>     >>> Regarding trunk, there seems to be many changes in trunk that
>>     break the
>>     >>> Windows MSVC compiler, it would take me a while to go through
>>     all of
>>     >>> those
>>     >>> changes (hmm I wonder why those changes made it through your CI
>>     >>> tests / why
>>     >>> they were merged before fixing the tests in Windows - as now
>>     it seems
>>     >>> like a
>>     >>> mountain of fixes are needed)
>>     >>
>>     >> I don't know or care about either cmake or Windows, so I don't
>>     have much
>>     >> invested in this. The Windows CI started failing in early May
>>     and I'm
>>     >> pretty sure it was not the relevant commit to trunk which triggered
>>     >> that, but an image change:
>>     >>
>>     >> https://github.com/apache/httpd/actions/workflows/windows.yml?
>>     >> query=branch%3Atrunk
>>     >>
>>     >> Regards, Joe
>>     >>
>>
>>     New pull request (into branch 2.4.x) :
>>     https://github.com/apache/httpd/pull/657
>>
>>     (backports earlier change, and also tackles the error :
>>          Unexpected enablement value "i" for mod_authnz_ldap
>>
>>
>>     thanks,
>>
>>     -jeff
>>
>>
>>
>>     -- 
>>     Jeff McKenna
>>     GatewayGeo: Developers of MS4W, & offering MapServer Consulting/Dev
>>     co-founder of FOSS4G
>>     http://gatewaygeo.com/
>>


-- 
Jeff McKenna
GatewayGeo: Developers of MS4W, & offering MapServer Consulting/Dev
co-founder of FOSS4G
http://gatewaygeo.com/