Windows - Install-Msys2.ps1 - minor reliability, taskkill, logging #928
Conversation
|
/azp run windows2016, windows2019 |
|
Azure Pipelines successfully started running 2 pipeline(s). |
|
JFYI, Windows 2019 has C:\Windows\System32\tar.exe, Windows 2016 does not. When 2016 is dropped, code can be revised to remove dependency on Git tar. Seems messy to include conditional code for it now... |
|
Ok, I'm confused. Azure Pipelines had all sorts of errors shown for the MSYS2 script. I just ran the script in Actions (copied the ImageHelpers folder also) and it installed fine: I haven't used Azure Pipelines for a while, so ? Happy to change whatever is needed... |
|
Hi @MSP-Greg , the reason for the fails is this line: This dash is incorrect symbol. |
|
Apologies, encoding seemed suspect, but I didn't think Actions and Pipelines would be using different encodings. I recall wrestling with it on AppVeyor... Fixed with a standard dash. |
|
/azp run windows2016, windows2019 |
|
Azure Pipelines successfully started running 2 pipeline(s). |
|
It appears that the last CI failed before it got to Install-Msys2.ps1 ? |
|
I pushed a change, adding the ~ folder ( |
Please revert it. We can't preserver user name for different environments. |
Please reconsider.
If you're adamant, I'll remove. Thanks. |
|
@MSP-Greg, After completion image generation all users are delete from image. |
|
@al-cheb miketimofeev Ok. Removed the code. |
| pacman.exe -Syyuu --noconfirm | ||
| pacman.exe -Syuu --noconfirm | ||
| taskkill /f /fi "MODULES eq msys-2.0.dll" |
There was a problem hiding this comment.
is this really necessary since we're installing?
There was a problem hiding this comment.
@alepauly, see #906 (comment). The procedure in this PR is different from the recommended in those references (-Syyuu vs -Syuu, and -Syuu vs -Suu). However, I think that this first execution of taskkill /f /fi "MODULES eq msys-2.0.dll" is required. I'm not sure about others, tho.
There was a problem hiding this comment.
I thought I'd err on the cautious side, as the script should be bullet-proof. I intend to keep an eye on future CI logs to see if all the taskkill statements are needed. Some were included based on doing the install in Actions (mult vm's on one system), which may not be the case for the Pipeliines runs...
There was a problem hiding this comment.
@MSP-Greg, I think that the first y is irrelevant (i.e. Syyuu or Syuu). However, I believe that using y in following statements is what might make taskkill to be required again. Anyway, this should be fixed upstream, albeit "experimental" (https://github.com/msys2/msys2-installer/releases/tag/2020-05-22). We should not need it in the near future.
There was a problem hiding this comment.
Anyway, this should be fixed upstream
As mentioned above, 'the script should be bullet-proof'. If an extra few seconds are added to the install time to do that, so be it. As MSYS2 is tested better re scripted install, etc, the install script may be able to be simplified.
Both Ruby itself and several extension gems are running CI using the new MSYS2 install, and all are passing with no changes. All are using the gcc tools, the bash shell is used less frequently.
There was a problem hiding this comment.
Both Ruby itself and several extension gems are running CI using the new MSYS2 install, and all are passing with no changes. All are using the gcc tools, the bash shell is used less frequently.
In fact, most users of eine/setup-msys2 are using update: true. Hence, apart from the Ruby extensions gems, there is intensive testing going on. See msys2/setup-msys2#20. For example in ghdl/ghdl:actions nightly packages for MSYS2 are being built with setup-msys2.
|
Ref #355 |
|
/azp run windows2016, windows2019 |
|
Azure Pipelines successfully started running 2 pipeline(s). |
Description
New tool, Bug fixing, or Improvement? Improvement
MSYS2 - add small changes to make MSYS2 installation/update via script more stable.
Update logging code for both install/update and final package listing.
Tested both locally and with GHA