Segmentation fault in Windows 10

[galmiza]

Hello,

I compiled the source using msys2/mingw64 on Windows 10.
The execution stops just after P1 table is created with error: segmentation fault.
Data is successfully created in temp drive.

Below some details about the execution with gdb.

Working Directory:   f:/
Working Directory 2: f:/
Plot Name: plot-k32-2021-06-08-14-02-53a515aecd736d6f9ab0a2d3f7a12eb62a907d5416ceab1b87bfe5ab419d84cb
[New Thread 15752.0x4db8]
[New Thread 15752.0x3d00]
[New Thread 15752.0x2edc]
[New Thread 15752.0x36b0]
[New Thread 15752.0x26dc]
[New Thread 15752.0x380]
[New Thread 15752.0x6ec]
[New Thread 15752.0x3d34]
[New Thread 15752.0x384c]
[New Thread 15752.0x19a4]
[New Thread 15752.0x4448]
[New Thread 15752.0x3c8]
[P1] Table 1 took 67.874 sec
[Thread 15752.0x4e6c exited with code 0]
[Thread 15752.0x2cf4 exited with code 0]
[Thread 15752.0x3fa4 exited with code 0]
[Thread 15752.0x2edc exited with code 0]
[Thread 15752.0x26dc exited with code 0]
[Thread 15752.0x6ec exited with code 0]
[Thread 15752.0x3d34 exited with code 0]
[Thread 15752.0x384c exited with code 0]
[Thread 15752.0x19a4 exited with code 0]
[Thread 15752.0x4448 exited with code 0]
[Thread 15752.0x3c8 exited with code 0]
[Thread 15752.0x4db8 exited with code 0]
[Thread 15752.0x36b0 exited with code 0]
[Thread 15752.0x380 exited with code 0]
[Thread 15752.0x3d00 exited with code 0]
[New Thread 15752.0x3b2c]
[New Thread 15752.0x27a4]
[New Thread 15752.0x44e4]
[New Thread 15752.0x166c]
[New Thread 15752.0x271c]
[New Thread 15752.0x30b4]
[New Thread 15752.0x4b4]
[New Thread 15752.0x3ca4]
[New Thread 15752.0xe30]
[New Thread 15752.0xe9c]
[New Thread 15752.0x3568]
[New Thread 15752.0x4554]
[New Thread 15752.0x3ab0]
[New Thread 15752.0x2968]
[New Thread 15752.0x246c]
[New Thread 15752.0x1b88]
[New Thread 15752.0x4a0c]
[New Thread 15752.0x2130]
[New Thread 15752.0x1890]
[New Thread 15752.0x3e90]
[New Thread 15752.0x47a8]
[New Thread 15752.0x4e70]
[New Thread 15752.0x3350]
[New Thread 15752.0x3ec8]
[New Thread 15752.0x2b40]
[New Thread 15752.0x3288]
[New Thread 15752.0x4b34]
[New Thread 15752.0xca4]
[New Thread 15752.0x1e18]
[New Thread 15752.0xabc]

Thread 23 received signal SIGSEGV, Segmentation fault.
[Switching to Thread 15752.0x4b4]
0x00007ff75387d080 in blake3_compress_xof_sse41 ()
(gdb)

[sigi77777]

please do not post any exe, in this way we can prevent the new plotter from spreading accross non technical people. (p.s: I dont have it yet as well, but i am working on building it for windows.)

[twig123]

I compiled the source using msys2/mingw64 on Windows 10.

Could you provide details on how you ran the build? I've tried with Visual Studio and MSYS2... but obviously I'm missing information as I can't it configured correctly.

[galmiza]

Well, I actually did horrible changes a bit everywhere in various files from the git project.
I didn't even properly track all as it would slow me down a LOT.
The idea is to retro-document from a working solution if I find one.

For instance, I

• disabled completely sodium which always return compilation errors.
  -uninstalled sodium from msys2
  -removed all references to sodium (by rewritting some preprocessor options)
  -rewrote the randombytes_buf function (which was based on sodium)
• added and compiled uint128_t from its git (as for some reasons it wasn't accepted)
• injected everywhere I could preprocessor options to tell the compiler the target is Windows (as it seems not to detect it)
• ... and counting

Yup, it is a real mess, I'm not an expert in Windows or compiler at all.
Every micro step towards a solution leads to dozens of new problems. It seems it is never ending.

Edit: Obvioulsy the target is to minimize the changes and even undo some I listed above.

[madMAx43v3r]

That's some dedication, but keep in mind any replacement of randombytes_buf needs to be cryptographically safe.

[galmiza]

Very right. Replacements are dangerous.
Yeah I tried quite a few things guided by the errors I faced. No real strategy, just a few brute force tries :D
Thanks for spending time on that project, the little I could see is already impressive!