This makes ddb show files more descriptive.
Details
Diff Detail
- Lint
Lint Passed - Unit
No Test Coverage - Build Status
Buildable 9796 Build 10231: arc lint + arc unit
Event Timeline
sys/kern/kern_descrip.c | ||
---|---|---|
3847 | Maybe "levt"? (We're limited to four characters, right?) |
sys/kern/kern_descrip.c | ||
---|---|---|
3847 | I don't have strong feelings about the actual names so I'm happy to change it to whatever is agreeable. Yes, four characters according to the db_print_file format (which looks like it currently adds up to 77 chars) ... uh, although, coming back to actually look at the output, it looks like it's already a bit screwed up, probably due to 4-byte vs 8-byte pointers. db> show files File Type Data Flag GCFl Count MCount Vnode FPID FCmd fffff803810a5d10 pipe fffff802b88f98e8 00000003 0000 1 0 0 98244 sh fffff801631529a0 pipe fffff8024a193460 00000003 0000 2 0 0 98244 sh fffff801631529a0 pipe fffff8024a193460 00000003 0000 2 0 0 98244 sh fffff80163152aa8 vnod fffff8004c508ac8 00000001 0000 1 0 fffff8004c508ac8 98244 sh Looks like we could do more here if there's appetite for more refactoring. I don't think e.g. six characters for the type would kill us. |
sys/kern/kern_descrip.c | ||
---|---|---|
3847 | Here's what that might look like: https://github.com/rlibby/freebsd/commit/4abf39f023b5b735047bf6f6d72fc4952351b955 And example output (amd64). I pasted together a few sections of the output to get a sampling of different types: db> show files File Type Data Flag GCFl Count MCount Vnode FPID FCmd fffff800048e6af0 vnode fffff80004084c00 00000003 0000 6 0 fffff80004f03938 722 zsh fffff800048e6af0 vnode fffff80004084c00 00000003 0000 6 0 fffff80004f03938 722 zsh fffff800048e6af0 vnode fffff80004084c00 00000003 0000 6 0 fffff80004f03938 722 zsh fffff80004920370 pipe fffff8000494fd48 00000003 0000 4 0 0 708 sleep fffff80004920370 pipe fffff8000494fd48 00000003 0000 4 0 0 708 sleep fffff80004907e60 pipe fffff8000494fbe0 00000003 0000 2 0 0 707 logger fffff800048d9820 vnode fffff80003dc9a00 00000003 0000 1 0 fffff800048d6b10 707 logger fffff80004920d20 socket fffff80004a78368 00000003 0000 1 0 0 321 dhclient fffff80004920cd0 pipe fffff8000497c8e8 00000003 0000 1 0 0 321 dhclient |
It adds two chars before p_comm.
For 4-byte pointers, it would be 67 up to p_comm.
For 8-byte pointers, it would be 91 up to p_comm.
Because this would also change the %12s (right justified) to %s for p_comm, it will be fewer characters when p_comm is less than 10 characters.
If we're spilling over 80 chars anyway, I don't see why shaving 2 chars off the type is helpful, so I prefer the new version. Approved, but perhaps give a couple of days for others to chime in.
Update patch to widen the ddb show files type field, and adjust
whitespace to line up columns on non-32-bit architectures.
Agreed. If we were still in 80 cols it would be good to not break that, but if we've already lost then oh well.