Closed
Description
Evaluate the effort and safety of a 32bit build of the installer.
Note: some preliminary findings out of a quick check:
- code-wise, a
s/-windows-x86_64/-windows-x86
in installer/src/Installer/Program.cs:56 would be enough to get a 32bit .msi; - the target installation should be changed from
Program Files
toProgram Files (x86)
[1]; - the installation eventually fails with error 13, potentially related to how the ODBC setup routines are called?
[1] The 32bit .msi find the 64bit driver and so it refuses to install -- not sure if the install path might be the only cause -- but this reveals that the name of the driver itself should bear the 32
tag (??).
The code itself can simply be made to work with 32-bit builders, resulting in a 32-bit installer (vs. having a 64-bit installer deliver both 64-bit and 32-bit drivers). Reasons for that could be:
- low effort of having a 32-bit installer (no significant changes required between the builds);
- low RM integration effort (besides the current
build.bat setup 64 package
we'd addbuild.bat setup 32 package
); - testing could be run similarly irrespective of underlying platform;
- the 32-bit installer could theoretically be used on 32-bit OSes (small chance of this occurrence, though);
- better user-experience: the user would need to make a choice right from the start and thus be made aware of the split (and of otherwise hidden potential issues later).
Metadata
Metadata
Assignees
Labels
No labels