> > Another question: Will Bernd worry about doslfn? > > Willi >
First, thank you Willi for initially discovering this problem. During the online-meeting, we discussed several things. But, nearly all of that time was spent on FDRepo and the DOSLFN problem. To remind everyone… When some packages are installed by FDNPKG under the FreeDOS kernel when the DOSLFN driver is loaded, the some of the files created by the install are damaged and the filesystem is corrupted. This occurs regardless of the type of FAT and even if the package does not contain any files that require LFN support. It appears to be connected to the usage of rare but legal characters in 8.3 filenames. For example, packages which contain an exclamation point or ampersand seem to trigger this bug consistently. The bug is known to occur and can be easily trigger by installing the current version the “ARACHNE” or similar packages using FreeDOS_Kernel+DOSLFN+FDNPKG. If any one of those 3 is changed, the bug does not appear. For example, EDR+DOSLFN+FDNPKG or FreeDOS+DOSLFN+FDINST will not exhibit the issue. This issue does not affect the FreeDOS installer or packages installed through FDIMPLES and FDINST. As a demonstration in a virtual machine, a fresh install of FreeDOS was checked with Microsoft ScanDisk. No errors were detected. This was follow by “fdnpkg install arachne.zip” which took a while to download and install. No error messages were displayed and the installation appeared to complete successfully. However, immediately running ScanDisk again showed multiple filesystem problems. A summary of what was discussed... There was a lot of concern expressed by everyone over this filesystem corruption bug. While it is possible to modify FDRepo to “quarantine” files that use uncommon legal 8.3 filename characters when generating a package, that would not actually fix the underlying problem. At this time, it is not known if this bug may occur elsewhere. For example, it is possible that a user saving a file that contains an explanation mark from a text editor may trigger the bug and corrupt their drive. The consensus was that the a test program should probably be created in attempt to try and replicate and detect the when exactly issue occurs. What is needed now is a volunteer to further investigate this issue and possibly create that test program. Do not look to me to create it or find the bug. I have far too many other things I must work on. We need to have someone step-up to try and fix this problem. Otherwise, we are likely to be stuck with this critical filesystem corruption bug for many years.
_______________________________________________ Freedos-devel mailing list Freedos-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/freedos-devel
