Skip to content
Merged
Changes from 1 commit
Commits
Show all changes
94 commits
Select commit Hold shift + click to select a range
707b995
enable streaming session retract tests
hnyls2002 Apr 13, 2026
40aacf7
debug: log per-req uncached + slot state
hnyls2002 Apr 13, 2026
6aaac8f
debug: log alloc_token_slots and match_prefix state
hnyls2002 Apr 13, 2026
2073f9f
fix retract leak: use full kv_committed_len as prefix for streaming s…
hnyls2002 Apr 13, 2026
cc627ea
debug: log avail_size at event loop stages
hnyls2002 Apr 13, 2026
58ca20f
track slot ownership via active_pool_idxs; drop is_active reliance
hnyls2002 Apr 13, 2026
003619f
fix double-free: update slot.kv_allocated_len in tail-free
hnyls2002 Apr 13, 2026
9163e8c
fix: clamp tail-free above cache_protected_len
hnyls2002 Apr 13, 2026
4461cbc
TestStreamingSessionAbortLeakRepro: inherit stdout/stderr instead of …
hnyls2002 Apr 13, 2026
4a9bcee
remove debug prints
hnyls2002 Apr 13, 2026
80542c7
TestStreamingSessionAbortLeakRepro: enable strict busy check
hnyls2002 Apr 13, 2026
8574e3d
Revert "TestStreamingSessionAbortLeakRepro: enable strict busy check"
hnyls2002 Apr 13, 2026
8641d49
Reapply "TestStreamingSessionAbortLeakRepro: enable strict busy check"
hnyls2002 Apr 13, 2026
56fa527
fix: clamp MatchResult.cache_protected_len to device_indices length
hnyls2002 Apr 13, 2026
0891517
fix: floor-align prefix_len to page_size in match_prefix
hnyls2002 Apr 13, 2026
593bd90
Revert "fix: floor-align prefix_len to page_size in match_prefix"
hnyls2002 Apr 13, 2026
f296b52
Revert "fix: clamp MatchResult.cache_protected_len to device_indices …
hnyls2002 Apr 13, 2026
fb2f7b4
Reapply "fix: clamp MatchResult.cache_protected_len to device_indices…
hnyls2002 Apr 13, 2026
6a428ef
Reapply "fix: floor-align prefix_len to page_size in match_prefix"
hnyls2002 Apr 13, 2026
2383aa2
debug: log tree state before/after cache_finished_req
hnyls2002 Apr 13, 2026
a7e02f9
debug: log session_held_tokens slot state
hnyls2002 Apr 13, 2026
6bcbf22
shrink: clamp slot.cache_protected_len by req.cache_protected_len
hnyls2002 Apr 13, 2026
580b0e6
remove debug logs
hnyls2002 Apr 13, 2026
cb73898
Revert "TestStreamingSessionAbortLeakRepro: inherit stdout/stderr ins…
hnyls2002 Apr 13, 2026
c7e338a
Merge remote-tracking branch 'origin/main' into lsyin/enable-streamin…
hnyls2002 Apr 13, 2026
515d5bd
add eagle streaming session test; allow -1 inherit offset for spec
hnyls2002 Apr 13, 2026
dc7e790
clarify kv_inherit_offset comment: overlap defers last commit
hnyls2002 Apr 13, 2026
6fd797f
add eagle spec v2 streaming session test; remove spec v2 gate
hnyls2002 Apr 13, 2026
c039eeb
drop unused SessionSlot.is_active
hnyls2002 Apr 13, 2026
f06f512
clamp swa_evicted_seqlen in match_prefix tail-free
hnyls2002 Apr 13, 2026
975a30d
add streaming session SWA tests (gpt-oss-20b)
hnyls2002 Apr 13, 2026
0d0732e
delete dead rematch path in release_session
hnyls2002 Apr 13, 2026
f9b112c
clamp req.swa_evicted_seqlen to prefix_len in shrink
hnyls2002 Apr 13, 2026
c0ba59f
un-skip SWA abort leak test
hnyls2002 Apr 13, 2026
18b154e
Merge remote-tracking branch 'origin/main' into lsyin/enable-streamin…
hnyls2002 Apr 14, 2026
c059a18
Merge remote-tracking branch 'origin/main' into lsyin/enable-streamin…
hnyls2002 Apr 14, 2026
91404e9
rename _alive_streaming_session_count to _streaming_session_count
hnyls2002 Apr 14, 2026
5308c80
use _is_streaming helper in release_kv_cache
hnyls2002 Apr 14, 2026
29fd84a
tiny
hnyls2002 Apr 14, 2026
f13b891
add eagle retract tests; consolidate concurrent test methods
hnyls2002 Apr 14, 2026
bce9652
keep prefix_len unaligned; align only free_start and cache_protected_len
hnyls2002 Apr 14, 2026
b7cefa9
fix page align: ceil-align free_start, set lengths to prefix_len
hnyls2002 Apr 14, 2026
b58fe97
always clamp slot/req lengths to prefix_len after restore
hnyls2002 Apr 14, 2026
3b79c36
fix page align: keep prefix_len unaligned; ceil-align result_protected
hnyls2002 Apr 14, 2026
5eabe62
rollback req_nodes on abort; remove shrink dead code
hnyls2002 Apr 14, 2026
3dd03af
add large page_size=256 streaming session test
hnyls2002 Apr 14, 2026
b46b2cd
move spec tail trim + flag sync into SessionAwareCache
hnyls2002 Apr 14, 2026
af64c27
simplify comment; move spec trim into SessionAwareCache
hnyls2002 Apr 14, 2026
351683b
keep tail-free in match_prefix; needed for retract alloc-commit gap
hnyls2002 Apr 14, 2026
a97b831
keep min(committed, token_ids); needed for retract retry logit reserve
hnyls2002 Apr 14, 2026
96c18db
remove _trim_speculative_tail; unify tail free in _free_tail
hnyls2002 Apr 14, 2026
c511de6
fix first-request abort: don't create slot for aborted first req
hnyls2002 Apr 14, 2026
bf0301f
fix mid-processing abort: check transient vs session pool_idx
hnyls2002 Apr 14, 2026
0900788
add first-request abort + mid-processing abort unit tests
hnyls2002 Apr 14, 2026
146f4aa
mid-processing abort: skip save_from_req to preserve slot state
hnyls2002 Apr 14, 2026
0d34575
simplify release_session docstring
hnyls2002 Apr 14, 2026
97d0da6
simplify abort rollback: peek + commit_req instead of pop + rollback
hnyls2002 Apr 14, 2026
63c595e
add mid-processing abort e2e test; use [x]=values() idiom
hnyls2002 Apr 14, 2026
ae17666
finish_req + inflight guard + mid-processing abort tail free + e2e test
hnyls2002 Apr 14, 2026
63ab282
ceil-align mid-processing abort tail free for paged allocator
hnyls2002 Apr 14, 2026
5b41cb8
fix abort test: kv_inherit_offset for Eagle; ceil-align tail free
hnyls2002 Apr 14, 2026
fbb77cb
abort = nuke all KV; inflight guard; EagleV2 skip; fix double-free
hnyls2002 Apr 14, 2026
f2ce814
merge origin/main; resolve conflicts with #22753 #22755
hnyls2002 Apr 15, 2026
e1a1037
pre-abort detach in match_prefix; ephemeral slot for first-mid; promp…
hnyls2002 Apr 15, 2026
7ec4c56
merge origin/main; abort PR #22790 landed
hnyls2002 Apr 15, 2026
76daf15
streaming session: fix retract tail leak via _free_tail in match_prefix
hnyls2002 Apr 15, 2026
955ffbd
merge lsyin/streaming-session-tail-leak (stack base)
hnyls2002 Apr 15, 2026
a1b4dda
ceil-align _free_tail start to page boundary to prevent paged allocat…
hnyls2002 Apr 15, 2026
8550db7
merge updated stack base (ceil-align fix)
hnyls2002 Apr 15, 2026
a4196c5
add page>1 + retract/spec test variants (RetractLargePage, EagleRetra…
hnyls2002 Apr 15, 2026
e797782
consolidate page>1 tests on page=256, raise chunked-prefill-size to 4096
hnyls2002 Apr 15, 2026
0ce9c40
simplify prefix_len comment
hnyls2002 Apr 15, 2026
b6801de
merge updated stack base (simpler comment)
hnyls2002 Apr 15, 2026
86cd6ab
shorten _free_tail docstring
hnyls2002 Apr 15, 2026
802c2de
merge updated stack base (shorter docstring)
hnyls2002 Apr 15, 2026
edd518c
clarify _free_tail docstring: gap from prefix not committed
hnyls2002 Apr 15, 2026
36ab6ae
merge stack base (clarify docstring)
hnyls2002 Apr 15, 2026
c9df814
drop cosmetic blank line in checker mixin
hnyls2002 Apr 15, 2026
b34236f
Merge branch 'main' into lsyin/enable-streaming-retract-tests
hnyls2002 Apr 15, 2026
8c50b18
simplify kv_cache_inheritance: drop post-close radix tree verificatio…
hnyls2002 Apr 15, 2026
1ff7eab
spec v2 pre-claim bonus slot in prepare; trim overshoot in cache_fini…
hnyls2002 Apr 15, 2026
bfd24e5
drop trim_overshoot from this PR (split to #22897)
hnyls2002 Apr 15, 2026
7a759f0
test: trim_overshoot must cap swa_evicted_seqlen
hnyls2002 Apr 15, 2026
4598dd3
rename trim_overshoot test to general postcondition (covers swa + nor…
hnyls2002 Apr 15, 2026
15c0de2
drop trim_overshoot unit test from this PR (split to #22900)
hnyls2002 Apr 15, 2026
245c74a
Merge remote-tracking branch 'origin/main' into lsyin/enable-streamin…
hnyls2002 Apr 15, 2026
6446aa5
drop redundant test variants: MixedChunk (covered by RetractMixedChun…
hnyls2002 Apr 15, 2026
df5971c
drop EagleRetract (covered by EagleRetractLargePage); upgrade EagleV2…
hnyls2002 Apr 15, 2026
a1624f0
drop standalone Retract; SWARetract -> SWARetractLargePage (page>1 is…
hnyls2002 Apr 15, 2026
c03eb48
drop stale module docstrings (just read the code)
hnyls2002 Apr 15, 2026
dd2aeaa
trim verbose docs in streaming session tests; keep inline section hea…
hnyls2002 Apr 15, 2026
02a3481
Merge branch 'main' into lsyin/enable-streaming-retract-tests
hnyls2002 Apr 15, 2026
8ac379c
update est_time from CI actuals (67->560, 420->450)
hnyls2002 Apr 15, 2026
8217254
Merge branch 'main' into lsyin/enable-streaming-retract-tests
hnyls2002 Apr 15, 2026
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 0 additions & 2 deletions test/registered/sessions/test_streaming_session.py
Original file line number Diff line number Diff line change
Expand Up @@ -621,7 +621,6 @@ def tearDownClass(cls):
kill_process_tree(cls.process.pid)


@unittest.skip("streaming session + retract has a token leak — tracked separately")
class TestStreamingSessionRetract(TestStreamingSession):
"""Streaming session under retract decode pressure."""

Expand Down Expand Up @@ -649,7 +648,6 @@ def tearDownClass(cls):
kill_process_tree(cls.process.pid)


@unittest.skip("streaming session + retract has a token leak — tracked separately")
class TestStreamingSessionRetractMixedChunk(TestStreamingSession):
"""Streaming session under retract decode with --enable-mixed-chunk."""

Expand Down
Loading