Installer updates

[MichaelHuth]

close #2206

If calling one of the powershell scripts fail the installer will now catch that and report through Messagebox (non-silent case).
In any case it will also return errorlevel code 1 then.

[MichaelHuth]

@t-b I noticed that we use the option to exit with an non-zero error level code only on one other location in the installer. I was pondering if we shouldn't actually add that to all cases where we quit early due to e.g. non-fulfilled requirements.

Currently this is only caught by the fact that an early abort leaves an incomplete installation. In this change the powershell scripts are called at the end, where it is not immediatly obvious that the installer quit early in silent mode.

[t-b]

@t-b I noticed that we use the option to exit with an non-zero error level code only on one other location in the installer. I was pondering if we shouldn't actually add that to all cases where we quit early due to e.g. non-fulfilled requirements.

Currently this is only caught by the fact that an early abort leaves an incomplete installation. In this change the powershell scripts are called at the end, where it is not immediatly obvious that the installer quit early in silent mode.

Yes let's do that.

[t-b]

Review:
f0843d7 (Update NSIS to version 3.10, 2025-03-05)
25c5c18 (Fixes in installer script for NSIS 3.10 update, 2025-03-05)
f2960b3 (Installer: Remove hint for MIES installation for 32-bit Igor Pro, 2025-03-06)
37789e0 (Installer: Bring installer to front after uninstaller ran, 2025-03-06)
2895e31 (Installer: Remove "Permanently remove MIES" MsgBox on uninstallation, 2025-03-06)
1775958 (Installer: Fix "uninstaller already running" error, 2025-03-06)
229a1c6 (Installer: Attempt admin uninstall only when actually admin, 2025-03-06)

Looks all good.

8c2db1c (Installer: Write out configuration to json file, 2025-03-06)

• How about creating the empty file on the fly in tools/create-installer.json?
• Can we have a short comment block in the code how the JSON looks like?
• The location of the installation_configuration.json should also be moved one level up next to version.txt.

bcf4eff (CI: Build release installer as unelevated version, 2025-03-07)

• While technically correct how about

        run: tools/create-installer.sh unelevated

instead?

• This change breaks the installer naming, as now also the user installer has the -cis suffix.

69062ad (Installer: Execute the ASRL disable script again, 2025-03-07)
8bac10b (Installer: Add handling of failed FixOffice365.ps1 call, 2025-03-10)
0e00a14 (Installer: Return well defined error code for install fail conditions, 2025-03-11)

Rest looks good.

• Please also adapt the documentation in Packages/doc/installation.rst.

[MichaelHuth]

@t-b

This change breaks the installer naming, as now also the user installer has the -cis suffix.

I can not reproduce this. The -cis naming is done in create-installer.sh and there was no change for this.
The unelevated installer still has the -cis suffix, whereas the elevated does not.

[t-b]

@t-b

This change breaks the installer naming, as now also the user installer has the -cis suffix.

I can not reproduce this. The -cis naming is done in create-installer.sh and there was no change for this. The unelevated installer still has the -cis suffix, whereas the elevated does not.

create-installer.sh has this code:

if [ "$#" -eq 0 ]
then
  echo "RequestExecutionLevel admin" > installer/$NSISREQUEST
else
  echo "RequestExecutionLevel user" > installer/$NSISREQUEST
  sed -i 's/\.exe/-cis.exe/' installer/$NSISOUTFILE
fi

so if I pass an argument I get an user installer but the name has the cis suffix.

[t-b]

AFAIU we can properly ditch the cis/dev installer as the only difference was the UAC elevation. Which we now don't do anymore in general.