HLS/previews: rerun on the CPU when the GPU can't be used right now

A busy GPU (llama-swap holding the 3090) made ffmpeg's CUDA setup fail
with CUDA_ERROR_OUT_OF_MEMORY, exit 187. The playlist generator took
that for a bad file and wrote playlist.unsupported, so the video never
played; preview clips were marked failed the same way.

is_gpu_init_failure() recognises CUDA/NVENC setup errors in ffmpeg's
stderr. When an NVENC run fails that way, the playlist and the preview
clip rerun once on libx264. A file that also fails on the CPU still gets
the sentinel. The HLS command and the timed run are pulled out of the
handler (hls_command, run_ffmpeg) so both attempts share them.

Fixes #215

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

src/video/ffmpeg.rs

@@ -36,6 +36,24 @@ pub async fn is_nvenc_available() -> bool {
     available
 }
 
+/// Whether ffmpeg's stderr says the GPU couldn't be used *right now*, as
+/// opposed to the input being bad (#215). `is_nvenc_available` only checks
+/// that ffmpeg was built with NVENC; on a GPU shared with llama-swap a resident
+/// model can leave no memory for a CUDA context, and ffmpeg then exits 187
+/// with `cuCtxCreate(...) failed -> CUDA_ERROR_OUT_OF_MEMORY`. Callers rerun
+/// such a failure on the CPU encoder instead of recording the file as broken.
+pub fn is_gpu_init_failure(stderr: &str) -> bool {
+    const MARKERS: &[&str] = &[
+        "CUDA_ERROR_",
+        "cuCtxCreate",
+        "OpenEncodeSessionEx failed",
+        "No capable devices found",
+        "Cannot load libcuda",
+        "Device creation failed",
+    ];
+    MARKERS.iter().any(|m| stderr.contains(m))
+}
+
 pub struct Ffmpeg;
 
 pub enum GifType {
@@ -271,25 +289,13 @@ fn parse_ffprobe_duration(stdout: &str) -> Option<f64> {
     None
 }
 
-/// Generate a preview clip from a video file.
-///
-/// Creates a ~10 second MP4 by extracting up to 10 equally-spaced 1-second segments
-/// at 480p with H.264 video and AAC audio. For short videos (<10s), uses fewer segments.
-/// For very short videos (<1s), transcodes the entire video.
-///
-/// Returns (duration_seconds, file_size_bytes) on success.
-pub async fn generate_preview_clip(input_file: &str, output_file: &str) -> Result<(f64, u64)> {
-    info!("Generating preview clip for: '{}'", input_file);
-    let start = Instant::now();
-
-    let duration = get_duration_seconds(input_file).await?;
-    let use_nvenc = is_nvenc_available().await;
-
-    // Create parent directories for output
-    if let Some(parent) = std::path::Path::new(output_file).parent() {
-        std::fs::create_dir_all(parent)?;
-    }
-
+/// The ffmpeg command for a preview clip, and the clip duration it produces.
+fn preview_clip_command(
+    input_file: &str,
+    output_file: &str,
+    duration: Option<f64>,
+    use_nvenc: bool,
+) -> (Command, f64) {
     let mut cmd = Command::new("ffmpeg");
     cmd.arg("-y");
 
@@ -350,7 +356,43 @@ pub async fn generate_preview_clip(input_file: &str, output_file: &str) -> Resul
     cmd.stdout(Stdio::null());
     cmd.stderr(Stdio::piped());
 
-    let output = cmd.output().await?;
+    (cmd, clip_duration)
+}
+
+/// Generate a preview clip from a video file.
+///
+/// Creates a ~10 second MP4 by extracting up to 10 equally-spaced 1-second segments
+/// at 480p with H.264 video and AAC audio. For short videos (<10s), uses fewer segments.
+/// For very short videos (<1s), transcodes the entire video.
+///
+/// Returns (duration_seconds, file_size_bytes) on success.
+pub async fn generate_preview_clip(input_file: &str, output_file: &str) -> Result<(f64, u64)> {
+    info!("Generating preview clip for: '{}'", input_file);
+    let start = Instant::now();
+
+    let duration = get_duration_seconds(input_file).await?;
+    let use_nvenc = is_nvenc_available().await;
+
+    // Create parent directories for output
+    if let Some(parent) = std::path::Path::new(output_file).parent() {
+        std::fs::create_dir_all(parent)?;
+    }
+
+    let (mut cmd, clip_duration) =
+        preview_clip_command(input_file, output_file, duration, use_nvenc);
+    let mut output = cmd.output().await?;
+
+    if use_nvenc
+        && !output.status.success()
+        && is_gpu_init_failure(&String::from_utf8_lossy(&output.stderr))
+    {
+        warn!(
+            "GPU unavailable for preview of '{}', retrying on the CPU encoder",
+            input_file
+        );
+        let (mut cmd, _) = preview_clip_command(input_file, output_file, duration, false);
+        output = cmd.output().await?;
+    }
 
     if !output.status.success() {
         let stderr = String::from_utf8_lossy(&output.stderr);
@@ -376,7 +418,28 @@ pub async fn generate_preview_clip(input_file: &str, output_file: &str) -> Resul
 
 #[cfg(test)]
 mod tests {
-    use super::parse_ffprobe_duration;
+    use super::{is_gpu_init_failure, parse_ffprobe_duration};
+
+    #[test]
+    fn gpu_busy_is_a_gpu_init_failure() {
+        // Verbatim from the e2e server with a llama-swap model resident (#215).
+        let oom = "[CUDA @ 0x5d1c2c0] cu->cuCtxCreate(&hwctx->cuda_ctx, desired_flags, hwctx->internal->cuda_device) failed -> CUDA_ERROR_OUT_OF_MEMORY: out of memory\nDevice creation failed: -542398533.";
+        assert!(is_gpu_init_failure(oom));
+        assert!(is_gpu_init_failure(
+            "[h264_nvenc @ 0x1] OpenEncodeSessionEx failed: out of memory (10): (no details)"
+        ));
+        assert!(is_gpu_init_failure(
+            "[h264_nvenc @ 0x1] No capable devices found"
+        ));
+    }
+
+    #[test]
+    fn bad_input_is_not_a_gpu_init_failure() {
+        assert!(!is_gpu_init_failure(
+            "[mov,mp4,m4a,3gp,3g2,mj2 @ 0x1] moov atom not found\nclip.mp4: Invalid data found when processing input"
+        ));
+        assert!(!is_gpu_init_failure("Conversion failed!"));
+    }
 
     #[test]
     fn empty_output_returns_none() {