From 8aa37bde3407a38f4d177ca561aae37da8d2baaa Mon Sep 17 00:00:00 2001
From: vjrj <vjrj@comunes.org>
Date: Wed, 22 Jul 2026 12:17:11 +0200
Subject: [PATCH] ci(release): retry network-fragile steps (fdroidserver
 download + fdroid build)

Even after the dnsmasq DNS cache, the Hetzner vSwitch still blips
mid-transfer. Two spots kept killing ~25min builds: the fdroidserver
tarball 'curl -sL | tar' (an empty stream -> 'gzip: unexpected end of
file') and maven/pub fetches mid-Gradle ('Network is unreachable').
Download fdroidserver to a file with curl --retry (+ -f + validate the
gzip before extract), and wrap fdroid build in a 3-attempt loop gated on
the APK actually appearing (fdroid build exits 0 even on failure).
---
 .forgejo/workflows/release.yml | 63 +++++++++++++++++++++++++++++-----
 1 file changed, 54 insertions(+), 9 deletions(-)

diff --git a/.forgejo/workflows/release.yml b/.forgejo/workflows/release.yml
index d0c9bcc..44575f8 100644
--- a/.forgejo/workflows/release.yml
+++ b/.forgejo/workflows/release.yml
@@ -150,8 +150,16 @@ jobs:
           # 2) fdroidserver from master, exactly like the fdroiddata CI.
           fds=/opt/fdroidserver-src
           mkdir -p "$fds"
-          curl -sL https://gitlab.com/fdroid/fdroidserver/-/archive/master/fdroidserver-master.tar.gz \
-            | tar -xz -C "$fds" --strip-components=1
+          # Download to a file with retries: a mid-transfer vSwitch blip used to
+          # pipe an empty stream straight into tar ("gzip: unexpected end of
+          # file"). -f fails on HTTP errors; --retry-all-errors covers resets.
+          for a in 1 2 3 4 5; do
+            curl -fsSL --retry 5 --retry-delay 3 --retry-all-errors \
+              https://gitlab.com/fdroid/fdroidserver/-/archive/master/fdroidserver-master.tar.gz \
+              -o /tmp/fds.tgz && tar -tzf /tmp/fds.tgz >/dev/null 2>&1 && break
+            echo "fdroidserver download attempt $a failed; retrying in 5s" >&2; sleep 5
+          done
+          tar -xz -C "$fds" --strip-components=1 -f /tmp/fds.tgz
           export PATH="$fds:$PATH"
           export PYTHONPATH="$fds:$fds/examples"
           git config --global --add safe.directory '*'
@@ -206,7 +214,14 @@ jobs:
           # per-ABI into its own job/matrix-entry (rather than all 3 sequentially in
           # one job) is what keeps each build inside the runner's max-job-runtime —
           # each needs a full cold NDK+tesseract-native+flutter compile.
-          fdroid build --verbose --no-tarball "org.comunes.tane:${vc}"
+          # Retry the build: a transient network blip mid-Gradle (maven/pub) used
+          # to kill it. `fdroid build` exits 0 even on failure, so gate the retry
+          # on the APK actually appearing. Up to 3 attempts fit the 90m timeout.
+          for a in 1 2 3; do
+            fdroid build --verbose --no-tarball "org.comunes.tane:${vc}" || true
+            if [ -f "unsigned/org.comunes.tane_${vc}.apk" ]; then break; fi
+            echo "fdroid build attempt $a produced no APK; retrying" >&2
+          done
 
           # `fdroid build` exits 0 even when builds fail — demand the APK.
           f="unsigned/org.comunes.tane_${vc}.apk"
@@ -281,8 +296,16 @@ jobs:
           . /etc/profile.d/bsenv.sh
           fds=/opt/fdroidserver-src
           mkdir -p "$fds"
-          curl -sL https://gitlab.com/fdroid/fdroidserver/-/archive/master/fdroidserver-master.tar.gz \
-            | tar -xz -C "$fds" --strip-components=1
+          # Download to a file with retries: a mid-transfer vSwitch blip used to
+          # pipe an empty stream straight into tar ("gzip: unexpected end of
+          # file"). -f fails on HTTP errors; --retry-all-errors covers resets.
+          for a in 1 2 3 4 5; do
+            curl -fsSL --retry 5 --retry-delay 3 --retry-all-errors \
+              https://gitlab.com/fdroid/fdroidserver/-/archive/master/fdroidserver-master.tar.gz \
+              -o /tmp/fds.tgz && tar -tzf /tmp/fds.tgz >/dev/null 2>&1 && break
+            echo "fdroidserver download attempt $a failed; retrying in 5s" >&2; sleep 5
+          done
+          tar -xz -C "$fds" --strip-components=1 -f /tmp/fds.tgz
           export PATH="$fds:$PATH"
           export PYTHONPATH="$fds:$fds/examples"
           git config --global --add safe.directory '*'
@@ -313,7 +336,14 @@ jobs:
           mkdir -p build
           git clone -q https://git.comunes.org/comunes/tane.git build/org.comunes.tane
           git -C build/org.comunes.tane checkout -q "${BUILD_REF}"
-          fdroid build --verbose --no-tarball "org.comunes.tane:${vc}"
+          # Retry the build: a transient network blip mid-Gradle (maven/pub) used
+          # to kill it. `fdroid build` exits 0 even on failure, so gate the retry
+          # on the APK actually appearing. Up to 3 attempts fit the 90m timeout.
+          for a in 1 2 3; do
+            fdroid build --verbose --no-tarball "org.comunes.tane:${vc}" || true
+            if [ -f "unsigned/org.comunes.tane_${vc}.apk" ]; then break; fi
+            echo "fdroid build attempt $a produced no APK; retrying" >&2
+          done
           f="unsigned/org.comunes.tane_${vc}.apk"
           if [ ! -f "$f" ]; then
             echo "expected $f, not found" >&2
@@ -382,8 +412,16 @@ jobs:
           . /etc/profile.d/bsenv.sh
           fds=/opt/fdroidserver-src
           mkdir -p "$fds"
-          curl -sL https://gitlab.com/fdroid/fdroidserver/-/archive/master/fdroidserver-master.tar.gz \
-            | tar -xz -C "$fds" --strip-components=1
+          # Download to a file with retries: a mid-transfer vSwitch blip used to
+          # pipe an empty stream straight into tar ("gzip: unexpected end of
+          # file"). -f fails on HTTP errors; --retry-all-errors covers resets.
+          for a in 1 2 3 4 5; do
+            curl -fsSL --retry 5 --retry-delay 3 --retry-all-errors \
+              https://gitlab.com/fdroid/fdroidserver/-/archive/master/fdroidserver-master.tar.gz \
+              -o /tmp/fds.tgz && tar -tzf /tmp/fds.tgz >/dev/null 2>&1 && break
+            echo "fdroidserver download attempt $a failed; retrying in 5s" >&2; sleep 5
+          done
+          tar -xz -C "$fds" --strip-components=1 -f /tmp/fds.tgz
           export PATH="$fds:$PATH"
           export PYTHONPATH="$fds:$fds/examples"
           git config --global --add safe.directory '*'
@@ -414,7 +452,14 @@ jobs:
           mkdir -p build
           git clone -q https://git.comunes.org/comunes/tane.git build/org.comunes.tane
           git -C build/org.comunes.tane checkout -q "${BUILD_REF}"
-          fdroid build --verbose --no-tarball "org.comunes.tane:${vc}"
+          # Retry the build: a transient network blip mid-Gradle (maven/pub) used
+          # to kill it. `fdroid build` exits 0 even on failure, so gate the retry
+          # on the APK actually appearing. Up to 3 attempts fit the 90m timeout.
+          for a in 1 2 3; do
+            fdroid build --verbose --no-tarball "org.comunes.tane:${vc}" || true
+            if [ -f "unsigned/org.comunes.tane_${vc}.apk" ]; then break; fi
+            echo "fdroid build attempt $a produced no APK; retrying" >&2
+          done
           f="unsigned/org.comunes.tane_${vc}.apk"
           if [ ! -f "$f" ]; then
             echo "expected $f, not found" >&2