Conversation
And sorted Skippedlines
A workaround of incorrect codegen of `SteamNetConnectionInfo_t` .
…-conditionalmarshal
…ps://github.com/Akarinnnnn/Steamworks.NET.AnyCPU into anycpu-packable-fix-align-by-conditionalmarshal
…-conditionalmarshal
Puropse is to exclude fake top level structs like CGameID::GameID_t. Maybe just add a table is enough?
|
Maybe it is too smart to my expectations for parser code. We just need to make it smart enough as git merge. I mean, some predefined special handling is enough just like old ways. |
|
Bad news here, I can't sync parser changes to my fork because git bugs. I did things below:
Would you guys meet this problem? This is very important for reviewing parser code changes. If you don't, would you help us to sync? I would gain write access of my fork to you. Additionally, my filesystem storing code is ReFS, working on NTFS might not meet this problem. |
|
@Chicken-Bones Incorrect marked pack aware structs should fixed a lot now. I've updated the gist to including the list. |
@Chicken-Bones Resolved, I pushed to GitHub directly on a NTFS drive, though some error messages thrown, it works. |
Simplify marshal fix #1 's review page, hide these `git subtree` imported commits from by merge them into base branch `anycpu-2`.
|
Reopen this, it's accidently closed by GitHub bots. |
Do we accept every time Steamworks SDK updates we regenerate special struct table by C++ compiler? Or we just using this method to debug marshal bugs when it happens again? |
…ps://github.com/Akarinnnnn/Steamworks.NET.AnyCPU into anycpu-packable-fix-align-by-conditionalmarshal
Our goal is to fix most callback system related marshal issues.
Added a conditional marshalling table to decide which alignment to use at runtime, this should fix most callback-retrieve related issues.
Struct parameters in interface methods are
stillneed to fix. Edit: Completed.I think we
stillneed some effort to fix parameters, maybe we still have to ship different assembly for different platforms.Edit: Parameter marshal is done, through it looks ugly.
Now the only problem is to find out which struct we treat it specially, in generation process.