nvenc_intra_refresh gets parsed into config::video.nv.intra_refresh but create_encoder() only checks client_config.enableIntraRefresh, so turning the setting on does nothing (Moonlight doesn't request intra refresh).
This checks both for H.264 and HEVC. Default is still off, and the capability check is unchanged, so encoders without intra refresh support just log the error like before.
Tested on Windows 11 build 26200 with an RTX 5060 Ti, using 2.0.1 plus this commit and Moonlight as the client. With nvenc_intra_refresh = enabled, the NvEnc: created encoder line for H.264 and HEVC includes intra-refresh, and AV1 is unchanged. With it disabled, the line matches stock 2.0.1. H.264 and HEVC streams looked normal both ways, with no dropped frames.
Activity
9Reviewed the intra-refresh configuration path. The driver-default retry has a status-reporting gap described below. The supplied hardware test covers the normal initialization path; I did not run a driver fallback test.
Retargeted to canary and rebased onto e3d13d1. The original commit applied with no conflicts. Built on canary with -DBUILD_TESTS=ON; test_fast_nvenc_log_flags and test_fast_nvenc_api pass.
Reviewed fd05111: no actionable correctness issues identified. The successful driver-default retry now reports intra-refresh as unknown when the explicit configuration requested it. I ran the four log-helper tests against the pinned NVENC headers; they passed. I did not run an encoder or hardware retry test.
Could you retarget this PR to the existing
canarybranch and resolve any resulting conflicts on your branch? That will make the changes reviewable against the intended integration base.