workflows/build.yml: Fix related to prolonged use of the msys2 job

[simbit18]

Summary

I am currently investigating the issue with the MSYS2 job.
To prevent the job from using the Windows Runner Image for an extended period (6 hours), the temporary solution is to:

• Use the previous setup-msys2 (GitHub Action to set up MSYS2).
  setup-msys2
• Add the maximum number of minutes that the msys2 job can run before GitHub automatically terminates it.
  GitHub Actions timeout

Impact

Impact on user: NO

Impact on build: Use of the Windows Runner image (Job Msys2) for max 60 minutes.

Impact on hardware: NO

Impact on documentation: NO

Impact on security: NO

Impact on compatibility: NO

Testing

GitHub

https://github.com/simbit18/manual-nuttx-ci/actions/runs/38146000862/job/114489527281#logs

[github-actions]

MemBrowse Memory Report

qemu-armv8a

• Data: .bss.g_rr_values +12 B (+0.0%, 77,474 B)
  No memory changes detected for:
• arduino-mega2560
• esp32-devkitc
• hifive1-revb
• mirtoo
• qemu-intel64
• rx65n-rsk2mb
• s698pm-dkit
• stm32-nucleo-f103rb

[casaroli]

Data from six recent msys2 runs, in case it helps: the hung ones (38116435520, 38116375368, 38117776252) and the good ones (38088288193, 38118616107, 38119950209) all used setup-msys2 ec48f7c5 and image windows-2025 20260901.588, so the pin may not be the cause. Every hung run stops in the first make olddefconfig, right after Select WINDOWS_MSYS=y, before the tools CMake prints TOOLS_DIR path is ... (about 15 s in a good run). The hangs are clustered in time (05:57–06:23 UTC on Oct 11; a run at 06:38 passed). The timeout is the right first step.

[simbit18]

Hi @casaroli Based on the tests I've been able to run in the limited time I have, the problem doesn't occur when build Nuttx with Make, but only with CMake.

It appears to be related to the updated version of the CMake package included in the latest version of MSYS2

New
20260927 > cmake-4.4.3-1

Old
20260611 > cmake-4.3.3-1

[casaroli]

The process dump from #20590's run (38162235073) shows where it hangs: under testbuild.sh only cmake -B build -DBOARD_CONFIG=nucleo-l152re/nsh -GNinja is left, with no child process. Its last output is Select WINDOWS_MSYS=y, and the next statement is nuttx_setconfig(), an execute_process(COMMAND setconfig ...) that captures the output. So setconfig has finished but CMake never gets the end of file on its pipe: the intermittent MSYS2/Cygwin pipe hang. A timeout around the configure in testbuild.sh would turn it into a failure that the existing retry ("Build Attempt N of 4") handles.