I'm not sure if I have the build script perfect, in that I'm not 100% that the files I want will be copied when the Github build action runs. I just don't have a great way to test the build process, but I am only changing two files here and I'm only adding, not subtracting.
File #1: If you look at the PSM1 file, you can see that I am just adding a check (when we are running windows) to see if we are x86 or x86. I then choose a path (sql-winclient-x86 or sql-winclient) based on that.
File #2: If you look at build-core.ps1, you can see that I am creating sql-winclient-x86 in a way that mimics what was already there for sql-winclient (IOW both windows and non-windows), then copying three files into sql-winclient-x86. Two of the new files are identical to the ones found in sql-winclient, but the third is copied from the 'native x86' location the in the nupkg (which was already cracked open by pre-existing code in the build script). Anything else in there is basically "she does this for win-sqlclient, so we probably need it for win-sqlclient-x86 too".
I did hand-build a version on my laptop ("2024.4.99", so I can test auto import). It works as I expected it to. So, "it works on my machine", but we all know how that goes. I will probably go ahead and deploy this in my estate on at least one production machine where I have a particularly thorny problem. When the official version gets released with a new version number, my version will be automatically superseded.
If I have this fix right, it closes #10.
I'm not sure if I have the build script perfect, in that I'm not 100% that the files I want will be copied when the Github build action runs. I just don't have a great way to test the build process, but I am only changing two files here and I'm only adding, not subtracting.
File #1: If you look at the PSM1 file, you can see that I am just adding a check (when we are running windows) to see if we are x86 or x86. I then choose a path (sql-winclient-x86 or sql-winclient) based on that.
File #2: If you look at build-core.ps1, you can see that I am creating sql-winclient-x86 in a way that mimics what was already there for sql-winclient (IOW both windows and non-windows), then copying three files into sql-winclient-x86. Two of the new files are identical to the ones found in sql-winclient, but the third is copied from the 'native x86' location the in the nupkg (which was already cracked open by pre-existing code in the build script). Anything else in there is basically "she does this for win-sqlclient, so we probably need it for win-sqlclient-x86 too".
I did hand-build a version on my laptop ("2024.4.99", so I can test auto import). It works as I expected it to. So, "it works on my machine", but we all know how that goes. I will probably go ahead and deploy this in my estate on at least one production machine where I have a particularly thorny problem. When the official version gets released with a new version number, my version will be automatically superseded.