May 2024 - wine-gitlab

[PATCH 0/3] MR5609: Rewrite of cmd command engine (part III).
by eric pouech (＠epo) 09 May '24

09 May '24

This is the third serie related to rewriting cmd's command engine. Currently, cmd stores block of commands as a linked list of CMD_LIST. Even for nested commands (like 'if' command inside a 'for' command), it's parsed as a single list of commands, using a 'depth' attribute and marker for end of block to describe the structure. This leads to several issues: - parsing for nested commands is not 100% robust, - chaining of commands (and its precedence) is not handled, - no separation between parsing and execution (esp. for 'IF' and 'FOR' commands, which behave slightly differently than regular commands when dealing with variables with delayed expansion), - data model doesn't 100% fit native behavior (eg redirections can apply to a block of commands, whereas redirections are attached to individual commands) - ... The final goal is to replace this with: - a tree of commands: + binary nodes for command chaining, + 'IF' and 'FOR' commands being parsed and stored as such + leaf being a command It will require a couple more series to reach that goal, so this series puts in place the basic building bits: - introducing a new structure CMD_NODE to hold that each node of such a tree, - moving the chaining operator into that CMD_NODE (meaning that we move from a linked list to a binary tree degenerated into a list: left is command, right is next node). -- https://gitlab.winehq.org/wine/wine/-/merge_requests/5609

2 3

[PATCH 0/8] MR5607: Link strmbase modules to msvcrt instead of ucrtbase.
by Elizabeth Figura (＠zfigura) 09 May '24

09 May '24

strmbase is built against msvcrt (and needs to remain as such to be used in tests). Hence modules that link to it also need to be built against msvcrt. In practice, this happens to work, but it causes some subtle strangeness. In particular, since ucrtbase does not export the simple printf functions, we end up instead linking to ntdll, whose implementation is notably incomplete, missing any float specifiers. While it's possible that ntdll should add those, we shouldn't be linking against ntdll's CRT anyway. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/5607

3 9

[PATCH 0/5] MR5604: Draft: winemac.drv: Some dock icon hiding improvements
by Tim Clem (＠tclem) 09 May '24

09 May '24

The most important thing here is a workaround for an OS bug: rapidly hiding and showing the dock icon can result in multiple icons for the same app, and those icons will ignore NSApp.applicationIconImage. Making sure transitions happen no less than 0.5 seconds apart from one another seems to fix it. Also, hiding a dock icon will deactivate an app. That's not desirable if it still has visible windows, so attempt to force a reactivation. That's often not successful on Sonoma due to the cooperative app activation heuristics, but it's the best we can do, and it should be a rare case. Also, there is some logic in -transformProcessToForeground: that we often want to happen regardless of whether the app is getting a dock icon (e.g. App Nap and activation). So I've renamed that function to more accurately communicate what it does, and consolidate it with -tryToActivateIgnoringOtherApps:. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/5604

3 6

[PATCH v2 0/5] MR5604: Draft: winemac.drv: Some dock icon hiding improvements
by Tim Clem (＠tclem) 08 May '24

08 May '24

The most important thing here is a workaround for an OS bug: rapidly hiding and showing the dock icon can result in multiple icons for the same app, and those icons will ignore NSApp.applicationIconImage. Making sure transitions happen no less than 1 second apart from one another seems to fix it. *Edit:* v2 update: 0.5 seconds is not always long enough. A good test case is the Notepad++ installer: after you select a language, it rapidly hides and shows some windows, and without a longer delay the icon will wind up with the default system terminal-looking icon. One second is probably overkill, but this ought to be a somewhat uncommon case. Also, hiding a dock icon will deactivate an app. That's not desirable if it still has visible windows, so attempt to force a reactivation. That's often not successful on Sonoma due to the cooperative app activation heuristics, but it's the best we can do, and it should be a rare case. Also, there is some logic in -transformProcessToForeground: that we often want to happen regardless of whether the app is getting a dock icon (e.g. App Nap and activation). So I've renamed that function to more accurately communicate what it does, and consolidate it with -tryToActivateIgnoringOtherApps:. -- v2: winemac.drv: Consolidate foregrounding and dock icon behavior. winemac.drv: Remove incorrect documentation for orderBelow: and rename a parameter. winemac.drv: Reactivate the app if needed after hiding its dock icon. winemac.drv: Enforce a delay between dock icon hides and shows. https://gitlab.winehq.org/wine/wine/-/merge_requests/5604

2 5

[PATCH 0/1] MR5602: user32/tests: Offset child CS_PARENTDC window from parent to make more bugs visible.
by Dmitry Timoshkov (＠dmitry) 08 May '24

08 May '24

-- https://gitlab.winehq.org/wine/wine/-/merge_requests/5602

2 1

[PATCH 0/4] MR5600: Second part of cmd engine rewrite.
by eric pouech (＠epo) 08 May '24

08 May '24

This is the second part of cmd's engine rewrite. It mainly contains code cleanup and simplifications. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/5600

2 4

[PATCH 0/5] MR5593: gdiplus: Start bracketing Graphics HDC uses.
by Esme Povirk (＠madewokherd) 07 May '24

07 May '24

This is the first part of a longer series of commits that can be found here: https://gitlab.winehq.org/madewokherd/wine/-/tree/getdc-wip?ref_type=heads The eventual goal is to no longer store gdi32 handles, which are a limited resource, on gdiplus objects, except when absolutely necessary. In particular, Graphics objects created from an HWND do not need to store an HDC (GetDC/ReleaseDC theoretically caches those and works just fine), and Bitmap objects with a gdi32-compatible format shouldn't store an HBITMAP or HDC. These changes revealed some other bugs in bitmap and linecap code, which I've also fixed in order to keep the tests working. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/5593

2 5

[PATCH 0/5] MR5592: Rewrite of cmd.exe's command engine (part I)
by eric pouech (＠epo) 07 May '24

07 May '24

This is the first serie of a major rewrite of cmd.exe's command engine (lexer, parser, command handling). The goal is to implement correctly: - command chaining (&&, ||) operators, - fix a number of issues (eg. 32679, 55401, 44063...) - remove some spaghetti code from engine, - allow to move forward on some pending MR (5246 5242 5194 4504...) This first part mainly extends the tests with the issues listed in above bug tickets & MR. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/5592

2 5

[PATCH 0/1] MR5591: winecrt0: Support __os_arm64x_dispatch_call symbol on ARM64EC.
by Jacek Caban (＠jacek) 07 May '24

07 May '24

This is needed for hybrid patchable guest exit thunks generated by the compiler. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/5591

2 1

[PATCH 0/1] MR5590: kernelbase: Add support for FileDispositionInfoEx to SetFileInformationByHandle().
by Dmitry Timoshkov (＠dmitry) 07 May '24

07 May '24

-- https://gitlab.winehq.org/wine/wine/-/merge_requests/5590

2 1

2025

2024

2023

2022

wine-gitlab May 2024