Hi, > I don't know to what extent these are i386-specific. If there's a > better place to discuss them, just let me know. the partitioning code is mostly machine-independent, though there are some exceptions (i386 is one of them). But I'd expect your issue to be in the mi code. > When partitioning the disk, I specifically switched the input unit to > sectors and specified an exact sector count. But then upon telling the > installer that I'm good with what it's got, it shows me the layout it's > chosen, and it's gone and silently changed the size - the size I > specified has been...well, not quite ignored, because the size chosen > is close to it, but it certainly hasn't been obeyed, even though the > previous display showed exactly what I entered. That's strange. I don't have that exactly in mind now, but there are some modifications made to the partition sizes you enter to obey sector sizes etc. > That's problem 1. Problem 2 arose because the partition in question is > at the end of the disk, and, when patching up the damage done by the > previous problem, I discovered that the partition editor provided at > that point has no way to say "change the size but leave the end fixed, > moving the start instead". The end isn't even modifiable, only the > beginning and the size, each leaving the other one fixed. There's also > no way to say "expand this partition to include the now-unused space > next to it"; I had to do the arithmetic to figure out the appropriate > start and size values. (Not that that's difficult, but it introduces > significant mistake potential when copying numbers around manually, and > it seems somewhat broken to have to use an external calculator, whether > electronic, cellulose and graphite, or wetware, when there's a > perfectly good computer right there.) Fixing that issue is some work, I think (see following text). > Problem 3 is comparatively minor: when editing a number (eg, when > patching up the damage done by problem 1), if the first character typed > is a backspace, it deletes not the last character, but the whole > number. This is extremely obnoxious when you want to delete just the > last character and (not unreasonably, I think) expect backspace to > delete just the previous character, the way it does just about > everywhere else, and unexpectedly lose the rest of the number. You're sure this is a sysinst issue and not msgc(1)? > If a PR would be a good idea, I can file one - or three, if that'd be > better. Imho, the whole code for the partitioning (mbr and disk label) of sysinst should be worked on. The current "two-phase" model (first setting size, then setting options for the FS) is unnecessary and overall the code for the partitioning is very complicated. I'll have to dig into the partitioning code soon again anyway, and a simple partitioning would make my work much better, so I could try to solve these issues and provide patches. But don't expect me to finish that in the next few weeks, so rather file a PR for that if you want to have it done reliably. Regards, Julian
Attachment:
signature.asc
Description: PGP signature