[shadye-6]

Summary

Fix matrix stack underflow in swPopMatrix() for all three matrix modes in src/external/rlsw.h. problem described in issue.

Closes #6361

Solution

changed the underflow checks for all three matrix stacks from <= 0 to <= 1, which prevents popping the initial matrix, sets SW_STACK_UNDERFLOW, and preserves the existing matrix stack state.

Testing

verified using a standalone rlsw test harness compiled with Clang 18.1.3, ASan and UBSan.

[shadye-6]

hey @raysan5, just following up on this PR, it's a pretty small fix, could you pls review? thanks for your time.

[raysan5]

@Bigfoot71 it looks good to me, despite the apparently AI generated issue, but please, could you take a second look to confirm merge?

[shadye-6]

hi @raysan5 , just following up again on this PR, it's a very small patch, and i've rewritten the issue as well in simpler terms. please can you have a look again. thanks for your time.

[raysan5]

@shadye-6 please, don't insist, PRs could take from weeks to months to be reviewed and merged, if merged. This is an open source project.

[Bigfoot71]

@raysan5 Thanks for the ping, reviewed on my end, the fix is correct, it can be merged.

---

Related note, I just noticed that swGetIntegerv() doesn't return the matrix stack depths:

raylib/src/external/rlsw.h

Lines 4078 to 4088 in 985ac29

	void swGetIntegerv(SWget name, int *v)
	{
	switch (name)
	{
	case SW_MODELVIEW_STACK_DEPTH: *v = SW_MODELVIEW_STACK_DEPTH; break;
	case SW_PROJECTION_STACK_DEPTH: *v = SW_PROJECTION_STACK_DEPTH; break;
	case SW_TEXTURE_STACK_DEPTH: *v = SW_TEXTURE_STACK_DEPTH; break;
	case SW_DRAW_FRAMEBUFFER_BINDING: *v = RLSW.boundFramebufferId; break;
	default: RLSW.errCode = SW_INVALID_ENUM; break;
	}
	}

It looks like a placeholder that was never completed, it should return the stack counter for each stack.
I can do another PR if you prefer.

[raysan5]

@Bigfoot71 thanks for the review! Merging!

I can do another PR if you prefer.

Yes, please.