https://1drv.ms/u/s!As-dKk-N73dSl2oMo-Zu9Ad0nTvO
MD5: dedf33040c888d3dd31bc8de5620af93

More logging for the problematic buffer. Also increased the buffer sizes by a factor of 4.

done

Thanks. Will check it out later.

Vogon's downloads are really slow maybe you should post the logs somewhere else. It takes like 5 minutes to download a file that's 900 KB.

https://1drv.ms/u/s!As-dKk-N73dSl2sMo-Zu9Ad0nTvO
MD5: 72c9c1e0669c6344c0d616fa86809621

The lockup appears to occur after some combination of VL_triangle and VL_tristrip, for which I didn't add logging earlier.

Let see if there're some nasties here.

https://fex.net/ru/s/2ol1ctb

sharangad wrote on 2024-02-24, 14:56:

https://1drv.ms/u/s!As-dKk-N73dSl2sMo-Zu9Ad0nTvO MD5: 72c9c1e0669c6344c0d616fa86809621 […]
Show full quote

https://1drv.ms/u/s!As-dKk-N73dSl2sMo-Zu9Ad0nTvO
MD5: 72c9c1e0669c6344c0d616fa86809621

The lockup appears to occur after some combination of VL_triangle and VL_tristrip, for which I didn't add logging earlier.

Let see if there're some nasties here.

https://fex.net/ru/s/doef9cn

That looks OK. I've disabled vsync as a test:
[EDIT]
Don't try this!
[EDIT2]
Try this
https://1drv.ms/u/s!As-dKk-N73dSl2wMo-Zu9Ad0nTvO
MD5: 39c1e8f2ed8c27f2cede75803f43bd27
[/EDIT2]
[/EDIT]

This should be good to test.

https://fex.net/ru/s/vczat8y

This is a bit worrying:

VL_InstallTextureMap_
Regular textures dims 128x64, stride 1
A 128x64 2 bytes per pixel texture should have a stride of 32 16-bt words. This says it's 2 bytes per row, unless it's always 1 for a non-paletted textrure.

The texture lookup is carried out by:
SrcBase + (Y * V_Linebytes(SrcStride))  (X * V_PIXEL_SIZE(SrcFmt)).
Linebytes is related to stride, once again not sure if that'll be low OR 64 (or 128).
SrcBase is a register which points to the start address in Verité RAM of the texture and (X,Y) are the texture coordinates (([0,texture width), [0, texture height)))

[EDIT] Don't test this unless you're bored. Regular RReady has a separate texture selecton call for F1. I'll try and implement that.[/EDIT]
https://1drv.ms/u/s!As-dKk-N73dSl20Mo-Zu9Ad0nTvO
MD5: d4fb2daf8f93624c5c72456fa83e8e71

Instead of VL_Triangle_ calling VL_Triangle in system Verite, this replaces the it with the implementation from RReady.

I don't think it'll make a difference. I've also replaced the import of V_GetBufferStride with V_GetBufferLinebytes. It'll still use the text stride, but the value reported will be the linebytes.

The stride of 1 first occurs in the buffer preceding the locked buffer.
BufferGroup's are opaque, so I can't mess with them.

Nm. I need to see the linebytes value. Please run this if you can.

i am not bored to test anything) no problem at all)

Well just reupload file, i downloaded and see logs here insted of binary dll's )

???

Hang on.

https://1drv.ms/u/s!As-dKk-N73dSl24Mo-Zu9Ad0nTvO
MD5: d6942417ac66b9de21a187d49db231f8

This should do the trick.

[EDIT] Vogon's downloads are full speed again. Not entirely sure what happended there.

The behavior changed, now instead of black screen, game crashes , and i can't see window, with read or write error of memory by instruction at adress.

Provide please this dll's without log, so it will work fast and from many attempts i will catch that window.

after i blindly pressed button on this window which behind game, i returns back to os and its works , no hangs.

https://youtu.be/lvCkWYErgDE

ok. Well the test was to see if it crashes to the desktop.

Give me 30 mins.

https://1drv.ms/u/s!As-dKk-N73dSl28Mo-Zu9Ad0nTvO
MD5: f37f427bd65972344aba2a881c94ad887d412194
and
https://1drv.ms/u/s!As-dKk-N73dSl3AMo-Zu9Ad0nTvO
MD5: f37f427bd65972344aba2a881c94ad887d412194

The nologging and logging are identetical binares which are all logging.

tried to start game trough quickrace, became black screen and hang as before.

Aaaargh! I need to slow down.

Did you run 005 and 006 through the exact same sequence?

In 006 the first triangle call is:
0:58:26.420 VL_Triangle_ vt 26, size (RReady) 32, Rendition 32, v1 006BEDE0, v2 006BEE00, v3 006BEE20

and in 005:
VL_Triangle_ vt 5, size (RReady) 24, Rendition 24, v1 006BF894, v2 006BF8AC, v3 006BF8C4

I would've expected the same vertex types (vt 5 and 26) for them both times, since it's pretty much drawing the same polygons.