Closed avodovnik closed 3 years ago
I believe we have two options, but both are contingent on us dropping PublicSign
:
1) We simply remove strong name, as it seems to be a source of problems 2) We use proper strong name signing for as long as we support the full framework.
https://github.com/microsoft/playwright-dotnet/pull/1572 - I've got a draft PR up, that uses the full snk
to sign the assemblies
Yup, that is best, just fully strong name since you have the private key.
After investigating https://github.com/microsoft/playwright-dotnet/issues/1176, and reading up on documentation I'm unconvinced there is a benefit in providing a strong name for our assemblies.
Especially considering we mainly target .NET Core (even though we support
netstandard2.0
we know the majority of users will be using core), where the docs state "For .NET Core, strong-named assemblies do not provide material benefits.".The other benefits/reasons would be (for .NET framework):
While this was considered for a moment, this would only be needed for internal usage (i.e. testing tools), and honestly, we'd be better staying away from this and implementing that differently.
We do not. In fact, it's unlikely to work correctly, due to the fact they'd both reference the same
.playwright
folder and then start crashing on the driver.Same point as above - it's unlikely we'd perform well, being installed in the GAC. Equally, the performance penalty of not doing this would be negligible compared to the fact we're actually launching a browser in the back - that is to say, this library is not performance critical enough to consider this a (significant) drawback.
~We do not have that requirement, nor do I foresee it happening.~ This might be the one reason we'd consider doing this "right" (i.e. not use
PublicSign
).