On Fri May 1 00:50:51 2026 +0000, Trent Waddington wrote:
Can I get the binary? Or where you got it, if I can go there to get it. not sure it'll help... I've got some weird things happening here... (like, depending on the path to the **same** image, there are variations in the result...). Until I figure out the cause of this (bad init somewhere, memory thrashing...) I'll assume culprit is by me, and not this MR
anyway a couple of items found while testing: * when native returns an error for an "incorrect" file, it's done before any callback is invoked; current code can report errors after first callback * cause of such failures are not fully covered; after a bit or toying this includes: * NumberOfRvaAndSize == 0 * one of the section has a raw size which isn't a multiple of 16 (and surprisingly it doesn't fail if it's not a multiple of FileHeader.FileAlignement) * (that seems to cover all the failures I had locally) * if one callback returns FALSE, then function terminates with last error set to ERROR_OPERATION_ABORTED * if the allocation of the NTheader/sections copy block fails, we likely want to return an OOM error (which is already set by HeapAlloc) rather than INVALID_PARAMETER * As you no longer change the initial mapping, FILE_MAP_COPY shall be replaced by FILE_MAP_READ -- https://gitlab.winehq.org/wine/wine/-/merge_requests/10295#note_139003