From 6b0a987a6a04c6f88ddc57fb944642a8754e50ed Mon Sep 17 00:00:00 2001
From: pat-s <patrick.schratz@gmail.com>
Date: Sun, 9 Aug 2026 16:25:13 +0000
Subject: [PATCH] fix(build): keep source fallbacks out of the S3 package cache

The cache handed to build_binary_package() as s3_package_cache was the raw
bucket listing, and the build list subtracted the same listing. Neither could
tell a binary from a package whose build failed and was published as its CRAN
source, so every source fallback read as "already built" and was skipped for
good - which is how alpine324 accumulated ~13.5k of them.

- drop objects the slot's index reports as served from source, which bincraft
  marks by leaving the Built stamp off
- build s3_dt from the filtered listing too, since it is subtracted from the
  build list and would otherwise exclude the packages that need building
- log how many objects were dropped

Archived objects have no index record and are kept: unknown means binary, never
"rebuild it". A slot last indexed by a bincraft that predates the Built change
stamps everything, so its cache is unchanged from before.
---
 local/packages-to-build.R | 59 +++++++++++++++++++++++++++++++++++----
 1 file changed, 53 insertions(+), 6 deletions(-)

diff --git a/local/packages-to-build.R b/local/packages-to-build.R
index 887952d..3b508fb 100644
--- a/local/packages-to-build.R
+++ b/local/packages-to-build.R
@@ -89,14 +89,61 @@ s3_pkgs <- s3fs::s3_dir_ls(
   recurse = TRUE
 )
 
-# Save the raw S3 file listing for the build step to use as s3_package_cache
+file_names <- basename(s3_pkgs)
+
+# An object occupying a key is not proof a binary was built: a package whose
+# build failed has its CRAN source published under exactly that name. Left in
+# the cache, `build_binary_package()` reads it as "already built" and skips the
+# package forever, which is how alpine324 accumulated ~13.5k source tarballs.
+#
+# bincraft stamps `Built` only on records it actually built, so the slot's own
+# index distinguishes them. A slot last indexed by a bincraft that predates that
+# fix stamps `Built` on everything, so the cache is then unchanged from before.
+# Archived objects have no index record and are kept: unknown means binary,
+# never "rebuild it".
+index_url <- sprintf(
+  "https://cran.rpkgs.com/%s/%s/latest/src/contrib/PACKAGES.gz",
+  arch,
+  codename
+)
+source_served <- tryCatch(
+  {
+    con_idx <- gzcon(url(index_url, open = "rb"))
+    on.exit(close(con_idx), add = TRUE)
+    idx <- read.dcf(con_idx, fields = c("Package", "Version", "Built"))
+    sprintf(
+      "%s_%s.tar.gz",
+      idx[is.na(idx[, "Built"]), "Package"],
+      idx[is.na(idx[, "Built"]), "Version"]
+    )
+  },
+  error = function(e) {
+    cat(sprintf(
+      "WARNING: could not read %s (%s); keeping the full S3 cache\n",
+      index_url,
+      conditionMessage(e)
+    ))
+    character(0)
+  }
+)
+
+binary_cache <- setdiff(file_names, source_served)
+cat(sprintf(
+  "S3 cache: %d objects, %d served as CRAN source, %d usable binaries\n",
+  length(file_names),
+  length(file_names) - length(binary_cache),
+  length(binary_cache)
+))
+
+# Save the S3 file listing for the build step to use as s3_package_cache.
 # This avoids loading s3fs/reticulate in the build container, saving memory for
 # the dependency-installer subprocesses
-saveRDS(basename(s3_pkgs), "/mnt/cache/packages/s3_cache.rds")
-
-file_names <- basename(s3_pkgs)
-matches <- regexec("^([A-Za-z0-9.]+)_([0-9][^/]*)\\.tar\\.gz$", file_names)
-parts <- regmatches(file_names, matches)
+saveRDS(binary_cache, "/mnt/cache/packages/s3_cache.rds")
+# Built from the filtered listing, not the raw one: `s3_dt` is subtracted from
+# the build list below, so a source fallback left in here would exclude the very
+# package that needs building.
+matches <- regexec("^([A-Za-z0-9.]+)_([0-9][^/]*)\\.tar\\.gz$", binary_cache)
+parts <- regmatches(binary_cache, matches)
 parts <- parts[sapply(parts, length) == 3]
 s3_dt <- data.table(
   Package = sapply(parts, `[`, 2),
-- 
2.54.0