- wine-gitlab - winehq.org

[PATCH v5 0/2] MR2215: winsock: Fix two connect() issues.
by Paul Gofman (＠gofman) 21 Feb '23

21 Feb '23

I was mainly after the issue fixed by the second patch but stomped on another one while testing that and that one looks even more unfortunate to me. 1. When connect() is used on blocking socket and fails for whatever reason, next connect() will make a new independent connection attempt (both on Windows and Unix). In non-blocking mode however (and Unix sockets are always in non-blocking mode in Wine) on Linux the next connect() will always return an error (unlike Windows which will make a new connect attempt if the previous one completed and failed). So in Wine now even for blocking socket connect() will fail if previous attempt failed and return WSAECONNABORTED. The error returned by Linux connect() will be ECONNABORTED if getsockopt(SO_ERROR) was called on the socket after connect completed with failure, and true connect() fail error if getsockopt wasn't called. 2. connect() on blocking socket times out after ~21sec on Windows and after ~130 sec on Linux. That is the case when peer doesn't actively reject connection but just never replies anything. That is, client sent SYN and never heard back. So much longer timeout leads to what is seen as hang on exit in one game, but probably more important that it affect much more in various apps provoking timeouts or huge delays not just on exit. I did not find a way to set connect() timeout for socket on Windows. That difference mostly corresponds to number of SYN send retries. That is 7 on Linux by default and 4 on Windows (the delay between attempts starts from 1s and is doubled on each next attempt). There is TCP_SYNCNT socket option (Linux specific now, unfortunately) which allows to set the number of retries. Setting that to 4 yields 32 sec timeout (vs 21 sec on Windows). The remaining difference is probably due to different amount of time waited after last SYN is sent. There is also TCP_USER_TIMEOUT parameter which allows to effectively set the delay exactly, but it might interfere with SO_KEEPALIVE somehow and I hope that 32 vs 21 is not that critical already as 130 vs 21 and probably we can avoid complicating the thing with it. -- v5: server: Set TCP SYN count on sockets. server: Retry socket connection on ECONNABORTED error. https://gitlab.winehq.org/wine/wine/-/merge_requests/2215

3 4

[PATCH 0/1] MR2242: win32u: Do not set last error in wait_message() for zero count.
by Paul Gofman (＠gofman) 21 Feb '23

21 Feb '23

Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=54405 Also fixes Assassin Creed hanging on start (for a similar reason: it doesn't expect a spurious -1 last error). There is already a fix attempt sent by Ivan Chikish in !2067 (https://gitlab.winehq.org/wine/wine/-/merge_requests/2067). But I suppose it may be fixed simpler like in this patch. wait_message() with zero count is only called from NtUserPeekMessage(). I think even if for some reason driver's MsgWaitForMultipleObjectsEx returns a real error code we still don't want to set last error here when called from PeekMessage(). -- https://gitlab.winehq.org/wine/wine/-/merge_requests/2242

3 6

[PATCH v12 0/11] MR66: vkd3d-shader/hlsl: Broaden resources support v2 (PART 1).
by Francisco Casas (＠fcasas) 21 Feb '23

21 Feb '23

First part of v2 of !27, which aims to: * Allow allocation of variables of complex types that contain both numerics and objects across multiple register sets (regsets). * Support the tex2D and tex3D intrinsics, inferring generic samplers dimension from usage, writing sampler declarations, and writing sample instructions. * Support for arrays of resources for both SM1 and SM4 (not to be confused with the resource-arrays of SM 5.1, which can have non-constant indexes). * Support for resources declared within structs. * Support for synthetic combined samplers for SM1 and synthetic separated samplers for SM4, considering that they can be arrays or members of structs. * Imitate the way the native compiler assigns the register indexes of the resources on allocation, which proved to be the most difficult thing. * Support for object components within complex input parameters. * Small fixes to corner cases. This part consists on parsing the `tex2D()` and `tex3D()` intrinsics and beginning to support the allocation of variables across multiple regsets. The whole series, is on my [master6](https://gitlab.winehq.org/fcasas/vkd3d/-/commits/master6) branch. -- v12: vkd3d-shader/hlsl: Allocate register reservations in a separate pass. vkd3d-shader/hlsl: Respect object reservations even if the object is unused. tests: Test allocation of unused objects. vkd3d-shader/hlsl: Allocate objects according to register set. vkd3d-shader/hlsl: Keep an hlsl_reg for each register set in hlsl_ir_var. vkd3d-shader/hlsl: Obtain extern resources as a separate array in SM4. https://gitlab.winehq.org/wine/vkd3d/-/merge_requests/66

5 20

[PATCH v4 0/2] MR2215: winsock: Fix two connect() issues.
by Paul Gofman (＠gofman) 21 Feb '23

21 Feb '23

I was mainly after the issue fixed by the second patch but stomped on another one while testing that and that one looks even more unfortunate to me. 1. When connect() is used on blocking socket and fails for whatever reason, next connect() will make a new independent connection attempt (both on Windows and Unix). In non-blocking mode however (and Unix sockets are always in non-blocking mode in Wine) on Linux the next connect() will always return an error (unlike Windows which will make a new connect attempt if the previous one completed and failed). So in Wine now even for blocking socket connect() will fail if previous attempt failed and return WSAECONNABORTED. The error returned by Linux connect() will be ECONNABORTED if getsockopt(SO_ERROR) was called on the socket after connect completed with failure, and true connect() fail error if getsockopt wasn't called. 2. connect() on blocking socket times out after ~21sec on Windows and after ~130 sec on Linux. That is the case when peer doesn't actively reject connection but just never replies anything. That is, client sent SYN and never heard back. So much longer timeout leads to what is seen as hang on exit in one game, but probably more important that it affect much more in various apps provoking timeouts or huge delays not just on exit. I did not find a way to set connect() timeout for socket on Windows. That difference mostly corresponds to number of SYN send retries. That is 7 on Linux by default and 4 on Windows (the delay between attempts starts from 1s and is doubled on each next attempt). There is TCP_SYNCNT socket option (Linux specific now, unfortunately) which allows to set the number of retries. Setting that to 4 yields 32 sec timeout (vs 21 sec on Windows). The remaining difference is probably due to different amount of time waited after last SYN is sent. There is also TCP_USER_TIMEOUT parameter which allows to effectively set the delay exactly, but it might interfere with SO_KEEPALIVE somehow and I hope that 32 vs 21 is not that critical already as 130 vs 21 and probably we can avoid complicating the thing with it. -- v4: server: Set TCP SYN count on sockets. server: Retry socket connection on ECONNABORTED error. https://gitlab.winehq.org/wine/wine/-/merge_requests/2215

4 5

[PATCH 0/1] MR2085: Add merge request template.
by Bernhard Kölbl (＠besentv) 20 Feb '23

20 Feb '23

Not sure if we want this, just wanted to bring the idea in. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/2085

3 3

Re: [PATCH v2 0/1] MR2236: gdi32/tests: Mark a couple other D3DKMTCheckOcclusion() tests as flaky. - approved
by Alexandre Julliard (＠julliard) 20 Feb '23

20 Feb '23

This merge request was approved by Alexandre Julliard. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/2236

1 0

Re: [PATCH v2 0/1] MR2239: include/dbghelp: Ensure SymRefreshModuleList() is properly defined. - approved
by Alexandre Julliard (＠julliard) 20 Feb '23

20 Feb '23

This merge request was approved by Alexandre Julliard. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/2239

1 0

Re: [PATCH v2 0/1] MR2238: dbghelp/tests: Test that overlapped module is really unloaded. - approved
by Alexandre Julliard (＠julliard) 20 Feb '23

20 Feb '23

This merge request was approved by Alexandre Julliard. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/2238

1 0

Re: [PATCH v5 0/4] MR2185: ntoskrnl: PnP device case. - approved
by Alexandre Julliard (＠julliard) 20 Feb '23

20 Feb '23

This merge request was approved by Alexandre Julliard. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/2185

1 0

Re: [PATCH v4 0/4] MR2224: include: Add dcomp.h. - approved
by Alexandre Julliard (＠julliard) 20 Feb '23

20 Feb '23

This merge request was approved by Alexandre Julliard. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/2224

1 0

2025

2024

2023

2022

wine-gitlab