Missing zh_TW docs in the latest 2025.3 beta

[andy72039]

Detailed description of the issue

The following localized docs for zh_TW were found missing in the latest beta and alpha branch.

• changes.html
• userGuide.html
• keyCommands.html

Beta version file name: nvda_snapshot_beta-52361,447bd88f

We haven't done and seen anything unexpected on Crowdin.

Why are the other templates not appropriate in this case?

This is for translation issue and should be addressed before RC.

Have you asked for advice on how to report this issue via a community discussion? If so, please link to the discussion

This has been reported to NVDA-translations group (link).

[keyang556]

The alpha-52452,c07d2614 (2026.1.0.52452) still missing these translation files.

[seanbudd]

I think there is a strange race condition going on with scons - the documentation is getting built after the signing request

Today's beta (52457,bcf61baa) contains the user guide but not changes:
https://github.com/nvaccess/nvda/actions/runs/17484962336/job/49662310624

signExecApi(["dist\uninstall.exe"], ["uninstaller\uninstGen.exe"])
PowerShell -File ci\scripts\sign.ps1 -ApiToken *** -FileToSign D:\a\nvda\nvda\dist\uninstall.exe
64 bit Processing build\arm64\interfaces\nvdaController\nvdaController.acf
nvdaController.acf
Install file: "build\arm64\client\nvdaController.h" as "extras\controllerClient\arm64\nvdaController.h"
cl /Fobuild\arm64\client\client.obj /c build\arm64\client\client.cpp /TP /EHsc /nologo /Od /std:c++20 /permissive- /MT /W3 /WX /DUNICODE /D_CRT_SECURE_NO_DEPRECATE /DLOGLEVEL=15 /D_WIN32_WINNT=_WIN32_WINNT_WINBLUE /DNOMINMAX /DNDEBUG /Iinclude /Iinclude\wil\include /Imiscdeps\include /Ibuild\arm64 /Z7
client.cpp
SHA256 hash: 2A5AF6E90E5B868F55392D6205BA774874FDA2B0393C53E30CD30F1B7B1AA4AB
Converting changes (lang='zh_TW') at user_docs\zh_TW\changes.md.sub to user_docs\zh_TW\changes.html

The previous beta (52361,447bd88f) contains neither and has this log

signExecApi(["dist\uninstall.exe"], ["uninstaller\uninstGen.exe"])
PowerShell -File ci\scripts\sign.ps1 -ApiToken *** -FileToSign D:\a\nvda\nvda\dist\uninstall.exe
Converting keyCommands (lang='zh_TW') at user_docs\zh_TW\userGuide.md.sub to user_docs\zh_TW\keyCommands.html
SHA256 hash: 62F6EBEB13F3F4613F640C308293396DB387B883A9434F867AA1A96583A34E5A
cl /Fobuild\arm64\client\nvdaController_C.obj /c build\arm64\client\nvdaController_C.c /nologo /Od /std:c++20 /permissive- /MT /W3 /WX /DUNICODE /D_CRT_SECURE_NO_DEPRECATE /DLOGLEVEL=15 /D_WIN32_WINNT=_WIN32_WINNT_WINBLUE /DNOMINMAX /DNDEBUG /Iinclude /Iinclude\wil\include /Imiscdeps\include /Ibuild\arm64 /Z7
nvdaController_C.c
Converting userGuide (lang='zh_TW') at user_docs\zh_TW\userGuide.md.sub to user_docs\zh_TW\userGuide.html
MIDL.EXE /nologo /x64 /server none /prefix all nvdaController_ /header build\x86_64\client\nvdaController.h /cstub build\x86_64\client\nvdaController_C.c build\x86_64\interfaces\nvdaController\nvdaController.idl /acf build\x86_64\interfaces\nvdaController\nvdaController.acf
64 bit Processing build\x86_64\interfaces\nvdaController\nvdaController.idl
nvdaController.idl
Converting changes (lang='zh_TW') at user_docs\zh_TW\changes.md.sub to user_docs\zh_TW\changes.html

We can linearize the builds to fix this

[seanbudd]

@LeonarddeR - do you have any thoughts on this given your experience with scons?

[andy72039]

Thanks for the clear explanation and support! Hopefully this issue can be resolved smoothly.

[andy72039]

Hi @seanbudd
Though the userGuide is successfully included in the latest beta: 52457,bcf61baa, however, We found some translation are not included. They seemed to be the strings updated after Sep 1, for example this one.
Is this intended? Could you help check it and make sure the latest updates on Crowdin will be covered in 2025.3 release. Thanks!

[michaelDCurran]

building in parallel has always been problematic and somewhat broken with scons, especially in regards to user documentation. However, there was always the assumption that official CI builds were always non-parallel (-j 1). This was certainly the case with appveyor. Is this not true with Github? At least until we know that parallel builds do work correctly (which requires a bit of work) Github builds absolutely do need to be non-parallel just like Appveyor was.

[michaelDCurran]

At very least 2025.3 should probably be non-parallel. then we can try for a proper fix for 2026.1.

[andy72039]

Thanks for the clarification. So if I understand correctly, with non-parallel for 2025.3, the updated docs should be included in the public release? Thanks!

[seanbudd]

Hi @andy72039 - we fetch translations from Crowdin once a week. We will be releasing 2025.3rc1 today or tomorrow and it should include the latest zh_TW translations.