Commit bbaf573
committed
Restore publisher streaming integration lost in APS rebase
Rebasing the APS OpenRTB stack onto main resolved publisher.rs against its
pre-streaming shape: the file reverted to the buffered design while
streaming_processor.rs kept the end-to-end machinery from #867, leaving
BodyStreamDecoder/BodyStreamEncoder and their helpers dead. With CI denying
warnings, every workflow failed at compile (struct BodyStreamDecoder is never
constructed, plus 8 sibling errors), so no tests ran on the branch.
Re-merge publisher.rs three-way (base: pre-streaming main, ours: branch tip,
theirs: main) so the file carries main's #867 async streaming and inline SSAT
rendering with the APS additions reapplied on top:
- build_bid_map keeps main's signature (settings, request_origin,
include_debug_bid), the zero-dimension w/h omission, and the always-inline
sanitized adm path, plus the APS renderer field and the renderer-aware
hb_adid fallback (cache_id, then renderer bid_id, then ad_id)
- the APS bid-map test moves to the new signature; the include_adm debug flag
it exercised is superseded by main's debug_bid design
Also update debug_metadata_preserves_malformed_and_error_responses for the
fail-closed debug-header allowlist from the APS security review: content-type
is preserved, retry-after is redacted. The compile failure had masked this
test regression since the allowlist landed.1 parent 0cb06ef commit bbaf573
2 files changed
Lines changed: 3440 additions & 206 deletions
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
1846 | 1846 | | |
1847 | 1847 | | |
1848 | 1848 | | |
| 1849 | + | |
1849 | 1850 | | |
1850 | 1851 | | |
1851 | 1852 | | |
| |||
1877 | 1878 | | |
1878 | 1879 | | |
1879 | 1880 | | |
1880 | | - | |
1881 | | - | |
| 1881 | + | |
| 1882 | + | |
| 1883 | + | |
| 1884 | + | |
| 1885 | + | |
| 1886 | + | |
| 1887 | + | |
| 1888 | + | |
| 1889 | + | |
| 1890 | + | |
| 1891 | + | |
1882 | 1892 | | |
1883 | 1893 | | |
1884 | 1894 | | |
| |||
0 commit comments