I noticed it on releases/0.4.10 with the RosBEWin2.1.6 GCC4.7.2 *rls* configuration:
C:/0410rls/reactos/dll/win32/shell32/dialogs/filetypes.cpp: In function 'BOOL FileTypesDlg_InsertToLV(HWND, LPCWSTR, INT, LPCWSTR)':
C:/0410rls/reactos/dll/win32/shell32/dialogs/filetypes.cpp:663:9: warning: variable 'iLargeImage' set but not used [-Wunused-but-set-variable]
C:/0410rls/reactos/dll/win32/shell32/dialogs/filetypes.cpp: In function 'BOOL EditTypeDlg_UpdateEntryIcon(HWND, PEDITTYPE_DIALOG, LPCWSTR, INT)':
C:/0410rls/reactos/dll/win32/shell32/dialogs/filetypes.cpp:1040:9: warning: unused variable 'iLargeImage' [-Wunused-variable]
But I do assume, that MSVC compilers would also complain about that in *rls* cfg.
Please notice that before 0.4.10-dev-202-g 698cbc6184 which did splitup and restructure the code
this function was placed within the file folder_options.cpp and was named
InsertFileType(HWND hDlgCtrl, WCHAR * szName, PINT iItem, WCHAR * szFile)
and not yet within file_types.cpp with the new name
FileTypesDlg_InsertToLV(HWND hListView, LPCWSTR szName, INT iItem, LPCWSTR szFile)
Back then it did not have the iLargeImage variable yet, and it also didn't warn upon rls-cfg-compilation.
Therefore 0.4.10-dev-202-g 698cbc6184 from (#582) is indeed the *guilty revision* (2018-06-06).
It was done in the very first commit of that PR 2fe0eab721
It's bad practice to move and refactor/extend code in functionality within the very same commit, as it makes reviewing the changes much harder.
That assert never fired for the last 6 years, therefore it would have been okay also to strip it together with the variable, but using DBG_UNREFERENCED_LOCAL_VARIABLE(iLargeImage); allows to keep the assert for master.