[INTERNAL] List of rufus-4.5 enhancements and fixes

[pbatard]

This topic is flagged as INTERNAL and is therefore NOT open for comments.

• [52ca798] Add dd image verification? → Dropped to focus on codename "Mad Max" instead
• Should we add a "Please wait" on close, while we wait for conflicting process thread to exit? → The hang doesn't seem to be due to the search for processes, and I can no longer see the issue now anyway
• [dd8314b] If we go Mad Max for this release, we should move the Check device for bad blocks under the drive section, remove the Use Rufus MBR with... option, and add the Mad Max option to the advanced drive options.
• [03d46b3] Silence the Could not get device number for device \\.\Z: [0x00000001] Incorrect function. messages
• Process search thread is not started! on 2361?!? → Not much I can do. User's machine could not start the process search thread, and without direct access to it, it's going to be difficult to figure out why...
• [026afa7, but really 0f23c47] Persistence not working with Mint (well, that's what you get when you force the use of symbolic links and therefore NTFS) Now also reported in No persistence with Linux Mint 21.3 Cinnamon 64-bit #2428.
• [ac9a3f4] Revoked bootloaders message does not display when using MBR (Bliss-v15.8.6-x86_64-OFFICIAL-gapps-20230703.iso)
• [1be7eaf] If you cancel the Fido script once (while it's working on a lengthy operation) you cannot launch it again.
• [39a5ae6] Read error: [0x0000001B] The drive cannot find the sector requested. from Restoring image creates raw partition instead of saved NTFS #2452 → Most likely from Windows somehow seeing a smaller-sized media
• [8738e7a] VHD/VHDX writing uses img_report.image_size which is fine for uncompressed VHD but incorrect for compressed VHDX and results in missing data! Instead we should use the size of the mounted VHD/VHDX drive we read from.
• Will need a translation round for Mad Max. Oh and we got a bunch of unsolicited translations (ru, ar, pt-BR) to deal with.
• [abc3312] What the hell is going on with the way we create files for NTFS and the AMI UEFI NTFS driver? → Well, the AMI NTFS driver sure is buggy!
• [5eae8a6] Do we want to move the Ctrl-Select capability to outside ISO extraction (e.g. DOS/UEFI:NTFS/NB)
• [5061af2] Add md5dsum_totalbytes for existing md5sum.txt files by loading the file in RAM and performing path lookups for computation. Also, to speed up lookups, start from last location.
• 64KB buffers still produce md5sum errors with the buggy AMI NTFS driver... → Well we tried various size buffers from 32 KB to 4 MB (which is what 7-zip uses) but always end up with checksum errors with the AMI driver. And before people complain "But extracting with 7-zip doesn't produce errors, so just fix your app!!!", there's only so far I am willing to go to work around what is clearly a bug in the AMI driver, and you are more than welcome to invest your own time to figure out what else we could do differently in Rufus to work around the issue, especially as https://github.com/pbatard/AmiNtfsBug demonstrates that you don't really have to do anything special to trigger the bug and that 7-zip not triggering the bug is the exception rather than the rule...
• [34e6e43] UEFI:NTFS 2.4 does not seem to replace the drive for MBR partitioned drives....
• diskcopy.dll is not deleted on Alt-D (and neither is the Rufus directory) → Can't see it any more. At any rate we use SHFileOperationW(FO_DELETE) on the directory so it's up to the Shell API to handle access issues.
• [d0bc050] "Checking for conflicting processes..." (MSG_278) doesn't display as it should when the user presses START. → Unsurprising, considering we defined the message but never used it anywhere...
• [180a617] It'd be preferable if the UEFI revocation notice appeared before the ISO → ESP notice
• [cba95f7] Do we want to switch our internal GRUB to 2.12?
• [6ac2dfb] Regression: temp_dir is not properly populated
• [4ec57dc] Regression: Could not read Grub version from 'boot/grub/i386-pc/normal.mod'/Could not detect Grub version with Ubuntu 24.04 with current. Works fine in 4.4.
• We always seem to timeout on the conflicting process search. Anything we can do about it? → Not if I no longer see the issue... Did Microsoft's last Windows update sort this out? Because I could consistently see it up to last patch tuesday, and now everything seems fine...
• [df9e333] md5sum bootloader doesn't install for Ubuntu 24.04 (but md5sum_totalbytes gets added) → This is due to the Ubuntu folks not having added /efi/boot/boot###.efi to their md5sum.txt, whereas we rely on md5sum.txt to detect them...
• [2a6df95] Aaaand Ubuntu changed their config file in 24.04 yet again, which BREAKS THE FRIGGING PERSISTENCE
• [83e0bda] is_in_md5sum() seems to be missing some files, such as ./dists/noble/Release.gpg from Ubuntu 24.04 → Due to is_in_md5sum() partial matching files such as ./dists/noble/Release against ./dists/noble/Release.pgp
• [pbatard/rufus-web@114b5bd] Need to add the Fedora tweaked GRUB BIOS symlinks for Fedora 40
• SHFileOperationW() produces an 0x000006BA: The RPC server is unavailable exception when using Alt-D. This seems to be a Windows issue, but is there anything we can do about it? → Not really, no. Trying to use TRY_AND_HANDLE() against the exception doesn't do anything as it seems to be a Windows kernel one. VS filters these exceptions out, probably because the Windows kernel generates too many of them.
• [2106be0] mt86plus_7.00_64.grub.iso produces an assertion error in iso9660_fs.c, but that wasn't the case in 4.4?!?
• [a50d390] I'm not sure we actually want to make the Enable runtime UEFI validation option persistent, as you then get it applied to UEFI Shell and stuff without realizing it... Maybe make it non persistent in this build and see if people complain?
• gdisk complains that The protective MBR's 0xEE partition is oversized! Auto-repairing. → Also happened with 4.4. Looks like Windows is unconditionally setting the protective GPT partition to 2 TB regardless of the actual disk size...
• [59fd550] We appear to have a major regression with Windows installs, where it fails at the "Copy Windows files" stage... → Oh wow, who would have thought that Windows would be so anal about seeing only one GPT partition with ESP type (if not, you get Copying Windows Files stuck at 0% before producing an error)...