lseek: let the host OS set lseek errors

[mailwl]

In game Timespinner, the code:

    public AudioEngine(string settingsFile, TimeSpan lookAheadTime, string rendererId)
    {
        using (Stream input = TitleContainer.OpenStream(settingsFile))
        {
            using (BinaryReader reader = new BinaryReader(input))
            {
                // skip
                uint prcCurveOffset = reader.ReadUInt32(); // <-- here 0xFFFFFFFFu
                // skip
                reader.BaseStream.Seek((long)prcCurveOffset, SeekOrigin.Begin); <-- (long)prcCurveOffset == 4294967295 
                for (int index1 = 0; (long)index1 < (long)prcCurveCount; ++index1)  // <-- prcCurveCount = 0, never executes
                {
                       ///skip
                }
            }
        }
    }

bad C# u32>s64 conversion leads to huge offset, and in C-code it compares to (s32)(-1), too, which not equal (s64)(-1) and then crash.
But, as you may see, in C# code, result of Seek() to not checked at all, and even not used, so, we can pass needed result/error to C-code.
I did some other changes, which made the game playable, but i do not know, which helps, will make pr for it later

[mailwl]

found where game is crashed - i occasionally set Log Filter to "*:Debug" - empty filter will game playable.
Looks like there is a place where "debug" log has malformed format string

[squidbus]

I don’t know if this is just how you worded it, but there shouldn’t be any “workaround for (specific engine)”. If the engine code works on real hardware, we should figure out why and implement the behavior as the real lseek does it, not add custom workarounds like this.

[mailwl]

i checked on usual .NET, the code
reader.BaseStream.Seek( (long)4294967295, SeekOrigin.Begin);
returns 4294967295 with no exception (the same with others offsets bigger than file itself)

[mailwl]

looks like on ubuntu lseek doesn't even set the error

wl@DESKTOP-0004IRP:~$ cat ./file.txt
abcd
wl@DESKTOP-0004IRP:~$ cat main.cpp
#include <unistd.h>
#include <fcntl.h>
#include <iostream>
#include <string.h>

int main() {
    int fd = open("./file.txt", O_RDONLY);
    auto off = lseek(fd, 4555555555, SEEK_SET);
    std::cout << "fd:" << fd << ", off: " << off << ", err: " << strerror(errno) << std::endl;
    close(fd);
}
wl@DESKTOP-0004IRP:~$ g++ main.cpp
wl@DESKTOP-0004IRP:~$ ./a.out
fd:3, off: 4555555555, err: Success
wl@DESKTOP-0004IRP:~$

[squidbus]

My comment was less about the outward behavior of the function and more what is going on inside of it to make that happen.

[chronicer]

[Kernel.Fs] file_system.cpp:179 sceKernelClose: Closing /app0/VEHICLES/COMPETITORSOUNDS.BNDL
[Kernel.Fs] file_system.cpp:61 sceKernelOpen: path = /app0/GAMELOGIC/GAMEPLAY.BNDL flags = 0x0 mode = 0
[Common.Filesystem] io_file.cpp:386 Seek: Seeking past the end of the file
[Kernel.Fs] file_system.cpp:313 sceKernelLseek: sceKernelLseek: failed to seek

Yes, that's not C#, but also seek problem in NFS Hot Pursuit Remastered. May you know something about this?

[mailwl]

with this PR NFS Hot Pursuit Remastered works, but i undestood, i made wrong implemetation. need to leave system seek function set errors by itself

[chronicer]

with this PR NFS Hot Pursuit Remastered works, but i undestood, i made wrong implemetation. need to leave system seek function set errors by itself

This error is displayed on all versions of the emulator, I just thought maybe you know something about this error)

[mailwl]

i am not sure, i made some random changes on main branch, including this PR, works well (not exactly, 30 fps on my old notebook with 2060), i will update this PR in one-two days and tell, if it enough

[chronicer]

i am not sure, i made some random changes on main branch, including this PR, works well (not exactly, 30 fps on my old notebook with 2060), i will update this PR in one-two days and tell, if it enough

Bugs in NFS HP Remastered...

В общем у меня баги какие-то в ней, может это из-за 1050 ti, может быть из-за той ошибки. В гонке верхняя половина не рендерится. С твоим билдом всё ок, это на всех сборках так. Вот я и решил спросить, может быть знаешь хоть что-то о ней. Хотя у знакомого тоже такая же ошибка, но у него 2070 и всё ок.

[mailwl]

к сожалению, не смогу подсказать, в графике я полный ноль, но shadPS4 выставляет очень высокие требования к видеокарте - Vulkan 1.4, с большим количеством расширений