Repository navigation
Mgrasfixes - #188
Merged
Merged
Mgrasfixes#188
Conversation
X paints a window's visible region with a clip ID (draw field 0x50) when the screen masks cannot hold its clip; the kernel has the window's GL drawing match it (pp1winmode bit 4 + id). Keep a 2-bit plane per pixel, write it under pp1winmode bits 11:10, and test it in visible(). Load pp1winmode from context image word 4 and CP_WINDOW, whose words 1 and 2 are winmode and pp1winmode (the kernel stores them as one doubleword, winmode high), not the other way round. Both sides added the clip-ID plane and the GL window's pp1winmode. Reconciled on the reading the sources support: - CIDmatch is bits 7:4, one bit per ID (the kernel's 1 << (id + 4), the diagnostics' MGRAS_CIDMATCH(cid) = cid << 4), and 0 disables the check (mgras_tport.c sets 0 for "CID disabled"). Not a value compare at bit 5, which would miss ID 3 (0x80) and make 0 mean "ID 0 only". - Clip-ID writes: two planes, enabled by pp1winmode bits 11:10 (the CIDmask field; X's 0xC00), rather than ColorMaskMSBs. - CP_WINDOW: window mode in word 1, pp1winmode in word 2 (the kernel's doubleword is winmode << 32 | pp1winmode). Kept from the PR: clip IDs in the state hash, pp1winmode in the GL summary, and its tests, adapted to the above. Co-Authored-By: iblowmymind <28228415+iblowmymind@users.noreply.github.com> Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The file manager scrolls its double-buffered 12-bit window (drawn with draw field 2) by reading it to the host with read field 1 and writing it back a line up. Read field 1 fell back to the drawing target, buffer A, so the scroll copied stale pixels over the icons. It is DRBpointers' second buffer. 4/4/4 host pixels read a component as c / 0x11 and wrote n * 0x11, so the background 0x502020 came back as 0x441111. A 12-bit visual's components are the top nibbles (as X's fills store them): read c >> 4, write n << 4, and the round trip keeps the pixel. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
libGLcore's __glMgrasDrawBuffer sends three words: the buffers drawn before an odd number of swaps, the buffers drawn after, and the count less one (GL_BACK is [4, 1, 0] in a 24-bit visual, [2, 1, 0] in a 12-bit one; GL_FRONT [1, 4] / [1, 2]; both [5, 5, 1] / [3, 3, 1]). Word 1 is not an absolute PP1 selector: taken as one, every double-buffered demo traced (atlantis, powerflip, solidview send [4, 1, 0]) would draw into buffer A whatever the swap state. Pick the word by the swap state; B is bit 1 or 2, both buffers draw field 3, overlay selectors (0x4N) and DRAW_NONE keep the previous commit's handling. Fixes 12-bit double-buffered GL_BACK too, which the one-word logic drew to the front. Also put the PR tests' CP_WINDOW tokens in the kernel's word order (window mode in word 1). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Expose write masks, viewport and scissor bounds, clear state, matrix stacks and matrices through native libGLcore state queries. Returning placeholder zeros caused Alias PowerAnimator to restore disabled drawing and empty viewports. Add a regression that queries and restores CI drawing state, then checks the resulting clipped pixels. Validated with 111 MGRAS tests and live PowerAnimator menus, four grids with axes, and Shelves displaying correctly.
The kernel's MgrasValidateBanks sends token 0x98 with the bank GL draws into next (the window's displayed bank xor 1; main buffers, then the second set) before every SCHEDULE_SWAP and on window validation, and stores it in a parked context's image (words 16-17, flag bit 26). Traced with Maya: [1, 1], swap, frame in B; [0, 0], swap, frame in A. The GE model counted swaps instead, which loses the phase for good on any swap it does not see. Take the bank from the token and the context image; swaps toggle only for contexts the kernel never named a bank for. A single-buffered window's missing second page now falls back to A. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
glCopyPixels draws its image back by DMA: pixel state 0xDA8 sets the
size, the RSS list the transfer mode, _WRITE_DMAGESETUP [0, ?, ?, 0x49D0]
arms glDrawPixels' routine, and one host DMA line carries every row.
Only the texture restore form (0x511A) was modelled, so the write half
was dropped ("host DMA start 0xa7 with no write transfer armed"). Maya
copies the front buffer to the back after a full redraw and afterwards
redraws only the lines that change, so its back buffer kept stale
pixels and every other frame showed them (brown viewport).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Closed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
fixed from iblowmymind with more on top of this.