5 Commits

Author SHA1 Message Date
10672f3816 all: cut bloat found in the pike audit
Dead, vestigial, and over-built code with no behavior change (except the
bool-vocabulary trim, a deliberate behavior tightening):

- httpdl: drop the vestigial sort+copy in validSegs (work-stealing was
  removed in 270812d, so segments always tile [0,total) ascending now);
  validate order in place, which also rejects a scrambled sidecar. Inline
  the endOff() accessor to s.end.
- cli: shrink the boolean vocabulary to true/false, which is all that's
  accepted. The invented yes/no/1/0/on/off spellings are gone, so e.g.
  --enable-dht=yes (or yes in got.conf) is now a usage error.
- download: remove the dead, never-called Stat.Done() method.
- progress: drop speed(), a pure alias of humanSize; call humanSize directly.
- main: replace the jobsHolder mutex box with atomic.Pointer[[]job] for the
  forced-exit session save; the capability stays, the lock and type go.
- bt: collapse isAddrInUse's X||X (missinggo.IsAddrInUse is that exact
  string match); drop the now-unused missinggo import (tidy -> indirect).
- option: note that prealloc and falloc are identical here.
2026-06-21 15:45:56 +09:00
508708b039 httpdl/cli: drop --auto-split; classify errors by type, not message text
Two Pike cleanups.

--auto-split was a third connection-count policy (beside -x and -s) that added
an extra knob rather than keeping the model minimal. Removing it also retires the
now-dead autoConns/maxAutoConns and the per-host transport cap that existed only
to scale for it; min(split, M*-x) is the sole policy again.

main's exit-code mapping fell back to substring-matching third-party error text
(strings.Contains "connection refused"/"timeout"/...), which rots when a
dependency rewords a message. The typed checks (errors.Is on the syscall errno,
*net.DNSError, net.Error.Timeout, context.DeadlineExceeded) already cover the
real stdlib errors -- verified end to end: refused -> 6, bad host -> 19. The two
cases that genuinely needed the text match are our own errors, now typed
sentinels: httpdl.ErrTimeout (idle timeout) and the bt metadata timeout wrapping
context.DeadlineExceeded.
2026-06-20 23:55:39 +09:00
ebc630b231 httpdl: drop work-stealing segment pool for one worker per segment
The pool let an idle worker steal the back half of a slower in-flight
segment, behind a mutex that also serialized every advance. At the default
-x 1 it never fired, and it carried the trickiest invariant in the tree
(minSplit >= readBuf keeps an owner's in-flight write out of the stolen tail).

makeSegments already caps the segment count at the connection count, so the
segments map one-to-one onto workers: give each its own goroutine, advance
written with a plain atomic add, and snapshot the fixed slice with no lock.
pool/newPool/acquire/steal/advance and the seg.owned field all go away.

Net -156 lines. Behaviour at the user's flags is unchanged, except a slow
mirror's tail segment is no longer rebalanced onto idle connections.
2026-06-20 23:43:15 +09:00
554beb6b48 httpdl: add segment work-stealing and --auto-split
an idle worker now steals the back half of the busiest in-flight segment
instead of exiting, so one slow segment no longer drains alone over a single
connection. control file persists stolen tails; resume re-tiles.

add --auto-split (opt-in): derive the connection count from file size (one
per min-split-size, up to 16), overriding -x/-s. default stays at 1
connection for a single server.
2026-06-19 14:01:59 +09:00
eb8d9b9460 first commit 2026-06-19 12:32:14 +09:00