[OS-BUILD PATCHv4 0/8] ark: rework changelog generation process during dist-release

[Herton R. Krzesinski (via Email Bridge)]

From: Herton R. Krzesinski on gitlab.com
Merge Request: https://gitlab.com/cki-project/kernel-ark/-/merge_requests/1980

This series try to improve the changelog generation process done
during a dist-release target run. To avoid doing hacks and having
to run setup-source/genspec.sh twice, this series splits the changelog
creation from genspec.sh and changes the Makefile to use the splitted
code.

Signed-off-by: Herton R. Krzesinski <herton(a)redhat.com&gt;

---
 redhat/self-test/2001-dist-release.bats |   1 -
 redhat/Makefile                         |  16 +++---
 redhat/genlog.sh                        |  63 +++++++++++++++++++++++++
 redhat/genspec.sh                       |  81 +-------------------------------
 4 files changed, 74 insertions(+), 87 deletions(-)

[OS-BUILD PATCHv4 1/8] redhat: split changelog generation from genspec.sh

[Herton R. Krzesinski (via Email Bridge)]

From: Herton R. Krzesinski <herton(a)redhat.com&gt;

redhat: split changelog generation from genspec.sh

This creates a new genlog.sh script to split out the changelog
generation code from genspec.sh. The splitted code is kept
as is.

Signed-off-by: Herton R. Krzesinski <herton(a)redhat.com&gt;

diff --git a/redhat/genlog.sh b/redhat/genlog.sh
new file mode 100755
index blahblah..blahblah 100755
--- /dev/null
+++ b/redhat/genlog.sh
@@ -0,0 +1,90 @@
+#!/bin/bash
+
+LAST_MARKER=$(cat "${REDHAT}"/marker)
+clogf="$SOURCES/changelog"
+# hide [redhat] entries from changelog
+HIDE_REDHAT=1;
+# hide entries for unsupported arches
+HIDE_UNSUPPORTED_ARCH=1;
+# override LC_TIME to avoid date conflicts when building the srpm
+LC_TIME=
+
+GIT_FORMAT="--format=- %s (%an)%n%N%n^^^NOTES-END^^^%n%b"
+GIT_NOTES="--notes=refs/notes/${RHEL_MAJOR}.${RHEL_MINOR}*"
+
+lasttag=$(git rev-list --first-parent --grep="^\[redhat\]
kernel-${SPECKVERSION}.${SPECKPATCHLEVEL}" --max-count=1 HEAD)
+# if we didn't find the proper tag, assume this is the first release
+if [[ -z $lasttag ]]; then
+    if [[ -z ${MARKER//[0-9a-f]/} ]]; then
+        # if we're doing an untagged release, just use the marker
+        echo "Using $MARKER"
+        lasttag=$MARKER
+    else
+	lasttag=$(git describe --match="$MARKER" --abbrev=0)
+    fi
+fi
+echo "Gathering new log entries since $lasttag"
+# master is expected to track mainline.
+
+cname="$(git var GIT_COMMITTER_IDENT |sed 's/>.*/>/')"
+cdate="$(LC_ALL=C date +"%a %b %d %Y")"
+cversion="[$BASEVERSION]";
+echo "* $cdate $cname $cversion" > "$clogf"
+
+git log --topo-order --no-merges -z "$GIT_NOTES" "$GIT_FORMAT" \
+	^"${UPSTREAM}" "$lasttag".. -- ':!/redhat/rhdocs' |
"${0%/*}"/genlog.py >> "$clogf"
+
+if [ "$HIDE_REDHAT" = "1" ]; then
+	grep -v -e "^- \[redhat\]" "$clogf" |
+		sed -e 's!\[Fedora\]!!g' > "$clogf.stripped"
+	cp "$clogf.stripped" "$clogf"
+fi
+
+if [ "$HIDE_UNSUPPORTED_ARCH" = "1" ]; then
+	grep -E -v "^-
\[(alpha|arc|arm|avr32|blackfin|c6x|cris|frv|h8300|hexagon|ia64|m32r|m68k|metag|microblaze|mips|mn10300|openrisc|parisc|score|sh|sparc|tile|um|unicore32|xtensa)\]"
"$clogf" > "$clogf.stripped"
+	cp "$clogf.stripped" "$clogf"
+fi
+
+# If the markers aren't the same then this a rebase.
+# This means we need to zap entries that are already present in the changelog.
+if [ "$MARKER" != "$LAST_MARKER" ]; then
+	# awk trick to get all unique lines
+	awk '!seen[$0]++' "$SOURCES/$SPECCHANGELOG" "$clogf" >
"$clogf.unique"
+	# sed trick to get the end of the changelog minus the line
+	sed -e '1,/# END OF CHANGELOG/ d' "$clogf.unique" >
"$clogf.tmp"
+	# Add an explicit entry to indicate a rebase.
+	echo "" > "$clogf"
+	echo -e "- $MARKER rebase" | cat "$clogf.tmp" - >>
"$clogf"
+	rm "$clogf.tmp" "$clogf.unique"
+fi
+
+# HACK temporary hack until single tree workflow
+# Don't reprint all the ark-patches again.
+if [ -n "$(git log --oneline --first-parent --grep="Merge ark patches"
"$lasttag"..)" ]; then
+	# Throw away the clogf and just print the summary merge
+	echo "" > "$clogf"
+	echo "- Merge ark-patches" >> "$clogf"
+fi
+
+# during rh-dist-git genspec runs again and generates empty changelog
+# create empty file to avoid adding extra header to changelog
+LENGTH=$(grep -c "^-" "$clogf" | awk '{print $1}')
+if [ "$LENGTH" = 0 ]; then
+	rm -f "$clogf"
+	touch "$clogf"
+fi
+
+cat "$clogf" "$SOURCES/$SPECCHANGELOG" > "$clogf.full"
+mv -f "$clogf.full" "$SOURCES/$SPECCHANGELOG"
+
+# genlog.py generates Resolves lines as well, strip these from RPM changelog
+grep -v -e "^Resolves: " "$SOURCES/$SPECCHANGELOG" >
"$clogf".stripped
+
+test -f "$SOURCES/$SPECFILE" &&
+	sed -i -e "
+	/%%SPECCHANGELOG%%/r $clogf.stripped
+	/%%SPECCHANGELOG%%/d" "$SOURCES/$SPECFILE"
+
+echo "MARKER is $MARKER"
+
+rm -f "$clogf"{,.stripped};
diff --git a/redhat/genspec.sh b/redhat/genspec.sh
index blahblah..blahblah 100755
--- a/redhat/genspec.sh
+++ b/redhat/genspec.sh
@@ -1,16 +1,6 @@
 #!/bin/bash
 # shellcheck disable=SC2153
 
-LAST_MARKER=$(cat "${REDHAT}"/marker)
-clogf="$SOURCES/changelog"
-# hide [redhat] entries from changelog
-HIDE_REDHAT=1;
-# hide entries for unsupported arches
-HIDE_UNSUPPORTED_ARCH=1;
-# override LC_TIME to avoid date conflicts when building the srpm
-LC_TIME=
-# STAMP=$(echo $MARKER | cut -f 1 -d '-' | sed -e "s/v//"); # unused
-
 UPSTREAM=$(git rev-parse -q --verify origin/"${UPSTREAM_BRANCH}" || \
 	   git rev-parse -q --verify "${UPSTREAM_BRANCH}")
 
@@ -97,83 +87,7 @@ if [ -n "$RHSELFTESTDATA" ]; then
 	exit 0
 fi
 
-GIT_FORMAT="--format=- %s (%an)%n%N%n^^^NOTES-END^^^%n%b"
-GIT_NOTES="--notes=refs/notes/${RHEL_MAJOR}.${RHEL_MINOR}*"
-
-lasttag=$(git rev-list --first-parent --grep="^\[redhat\]
kernel-${SPECKVERSION}.${SPECKPATCHLEVEL}" --max-count=1 HEAD)
-# if we didn't find the proper tag, assume this is the first release
-if [[ -z $lasttag ]]; then
-    if [[ -z ${MARKER//[0-9a-f]/} ]]; then
-        # if we're doing an untagged release, just use the marker
-        echo "Using $MARKER"
-        lasttag=$MARKER
-    else
-	lasttag=$(git describe --match="$MARKER" --abbrev=0)
-    fi
-fi
-echo "Gathering new log entries since $lasttag"
-# master is expected to track mainline.
-
-cname="$(git var GIT_COMMITTER_IDENT |sed 's/>.*/>/')"
-cdate="$(LC_ALL=C date +"%a %b %d %Y")"
-cversion="[$BASEVERSION]";
-echo "* $cdate $cname $cversion" > "$clogf"
-
-git log --topo-order --no-merges -z "$GIT_NOTES" "$GIT_FORMAT" \
-	^"${UPSTREAM}" "$lasttag".. -- ':!/redhat/rhdocs' |
"${0%/*}"/genlog.py >> "$clogf"
-
-if [ "$HIDE_REDHAT" = "1" ]; then
-	grep -v -e "^- \[redhat\]" "$clogf" |
-		sed -e 's!\[Fedora\]!!g' > "$clogf.stripped"
-	cp "$clogf.stripped" "$clogf"
-fi
-
-if [ "$HIDE_UNSUPPORTED_ARCH" = "1" ]; then
-	grep -E -v "^-
\[(alpha|arc|arm|avr32|blackfin|c6x|cris|frv|h8300|hexagon|ia64|m32r|m68k|metag|microblaze|mips|mn10300|openrisc|parisc|score|sh|sparc|tile|um|unicore32|xtensa)\]"
"$clogf" > "$clogf.stripped"
-	cp "$clogf.stripped" "$clogf"
-fi
-
-# If the markers aren't the same then this a rebase.
-# This means we need to zap entries that are already present in the changelog.
-if [ "$MARKER" != "$LAST_MARKER" ]; then
-	# awk trick to get all unique lines
-	awk '!seen[$0]++' "$SOURCES/$SPECCHANGELOG" "$clogf" >
"$clogf.unique"
-	# sed trick to get the end of the changelog minus the line
-	sed -e '1,/# END OF CHANGELOG/ d' "$clogf.unique" >
"$clogf.tmp"
-	# Add an explicit entry to indicate a rebase.
-	echo "" > "$clogf"
-	echo -e "- $MARKER rebase" | cat "$clogf.tmp" - >>
"$clogf"
-	rm "$clogf.tmp" "$clogf.unique"
-fi
-
-# HACK temporary hack until single tree workflow
-# Don't reprint all the ark-patches again.
-if [ -n "$(git log --oneline --first-parent --grep="Merge ark patches"
"$lasttag"..)" ]; then
-	# Throw away the clogf and just print the summary merge
-	echo "" > "$clogf"
-	echo "- Merge ark-patches" >> "$clogf"
-fi
-
-# during rh-dist-git genspec runs again and generates empty changelog
-# create empty file to avoid adding extra header to changelog
-LENGTH=$(grep -c "^-" "$clogf" | awk '{print $1}')
-if [ "$LENGTH" = 0 ]; then
-	rm -f "$clogf"
-	touch "$clogf"
-fi
-
-cat "$clogf" "$SOURCES/$SPECCHANGELOG" > "$clogf.full"
-mv -f "$clogf.full" "$SOURCES/$SPECCHANGELOG"
-
-# genlog.py generates Resolves lines as well, strip these from RPM changelog
-grep -v -e "^Resolves: " "$SOURCES/$SPECCHANGELOG" >
"$clogf".stripped
-
-test -f "$SOURCES/$SPECFILE" &&
-	sed -i -e "
-	/%%SPECCHANGELOG%%/r $clogf.stripped
-	/%%SPECCHANGELOG%%/d" "$SOURCES/$SPECFILE"
-
-echo "MARKER is $MARKER"
+"${0%/*}"/genlog.sh
 
 if [ "$DISTRO" == "fedora" ]; then
 	# The tarball in the SRPM contains only the upstream sources.
@@ -188,5 +102,3 @@ else
 	# compatibility is necessary.
 	touch "${SOURCES}/patch-${SPECKVERSION}.${SPECKPATCHLEVEL}"-redhat.patch
 fi
-
-rm -f "$clogf"{,.stripped};

--
https://gitlab.com/cki-project/kernel-ark/-/merge_requests/1980

[OS-BUILD PATCHv4 2/8] redhat: don't hardcode temporary changelog file

[Herton R. Krzesinski (via Email Bridge)]

From: Herton R. Krzesinski <herton(a)redhat.com&gt;

redhat: don't hardcode temporary changelog file

We always create a changelog file inside the sources directory, which
is temporary and always deleted at the end. Instead hardcoding the
temporary file name, create a real temporary file to be used with
genlog.sh script, given as its first argument.

Signed-off-by: Herton R. Krzesinski <herton(a)redhat.com&gt;

diff --git a/redhat/genlog.sh b/redhat/genlog.sh
index blahblah..blahblah 100755
--- a/redhat/genlog.sh
+++ b/redhat/genlog.sh
@@ -1,7 +1,7 @@
 #!/bin/bash
 
 LAST_MARKER=$(cat "${REDHAT}"/marker)
-clogf="$SOURCES/changelog"
+clogf="$1"
 # hide [redhat] entries from changelog
 HIDE_REDHAT=1;
 # hide entries for unsupported arches
@@ -70,8 +70,7 @@ fi
 # create empty file to avoid adding extra header to changelog
 LENGTH=$(grep -c "^-" "$clogf" | awk '{print $1}')
 if [ "$LENGTH" = 0 ]; then
-	rm -f "$clogf"
-	touch "$clogf"
+	echo -n > "$clogf"
 fi
 
 cat "$clogf" "$SOURCES/$SPECCHANGELOG" > "$clogf.full"
@@ -87,4 +86,4 @@ test -f "$SOURCES/$SPECFILE" &&
 
 echo "MARKER is $MARKER"
 
-rm -f "$clogf"{,.stripped};
+rm -f "$clogf".stripped;
diff --git a/redhat/genspec.sh b/redhat/genspec.sh
index blahblah..blahblah 100755
--- a/redhat/genspec.sh
+++ b/redhat/genspec.sh
@@ -87,7 +87,9 @@ if [ -n "$RHSELFTESTDATA" ]; then
 	exit 0
 fi
 
-"${0%/*}"/genlog.sh
+clogf=$(mktemp)
+trap 'rm -f "$clogf"' SIGHUP SIGINT SIGTERM EXIT
+"${0%/*}"/genlog.sh "$clogf"
 
 if [ "$DISTRO" == "fedora" ]; then
 	# The tarball in the SRPM contains only the upstream sources.

--
https://gitlab.com/cki-project/kernel-ark/-/merge_requests/1980

[OS-BUILD PATCHv4 3/8] redhat: drop merge ark patches hack

[Herton R. Krzesinski (via Email Bridge)]

From: Herton R. Krzesinski <herton(a)redhat.com&gt;

redhat: drop merge ark patches hack

Changelog generation carries a temporary hack to add a merge ark-patches
entry in the changelog. I think we can drop this today.

Signed-off-by: Herton R. Krzesinski <herton(a)redhat.com&gt;

diff --git a/redhat/genlog.sh b/redhat/genlog.sh
index blahblah..blahblah 100755
--- a/redhat/genlog.sh
+++ b/redhat/genlog.sh
@@ -58,14 +58,6 @@ if [ "$MARKER" != "$LAST_MARKER" ]; then
 	rm "$clogf.tmp" "$clogf.unique"
 fi
 
-# HACK temporary hack until single tree workflow
-# Don't reprint all the ark-patches again.
-if [ -n "$(git log --oneline --first-parent --grep="Merge ark patches"
"$lasttag"..)" ]; then
-	# Throw away the clogf and just print the summary merge
-	echo "" > "$clogf"
-	echo "- Merge ark-patches" >> "$clogf"
-fi
-
 # during rh-dist-git genspec runs again and generates empty changelog
 # create empty file to avoid adding extra header to changelog
 LENGTH=$(grep -c "^-" "$clogf" | awk '{print $1}')

--
https://gitlab.com/cki-project/kernel-ark/-/merge_requests/1980

[OS-BUILD PATCHv4 4/8] redhat: always add a rebase entry when ark merges from upstream

[Herton R. Krzesinski (via Email Bridge)]

From: Herton R. Krzesinski <herton(a)redhat.com&gt;

redhat: always add a rebase entry when ark merges from upstream

We have some complex logic to parse changelog entries to choose whether
we add a rebase entry or not to the changelog. Since genlog.py is adding
a Resolves entry always now, simply replace that logic with a
replacement using sed to add a rebase entry when the markers differ.
The marker check doesn't work at the moment due Makefile logic but
that will be fixed in the following changes.

Signed-off-by: Herton R. Krzesinski <herton(a)redhat.com&gt;

diff --git a/redhat/genlog.sh b/redhat/genlog.sh
index blahblah..blahblah 100755
--- a/redhat/genlog.sh
+++ b/redhat/genlog.sh
@@ -48,14 +48,9 @@ fi
 # If the markers aren't the same then this a rebase.
 # This means we need to zap entries that are already present in the changelog.
 if [ "$MARKER" != "$LAST_MARKER" ]; then
-	# awk trick to get all unique lines
-	awk '!seen[$0]++' "$SOURCES/$SPECCHANGELOG" "$clogf" >
"$clogf.unique"
-	# sed trick to get the end of the changelog minus the line
-	sed -e '1,/# END OF CHANGELOG/ d' "$clogf.unique" >
"$clogf.tmp"
-	# Add an explicit entry to indicate a rebase.
-	echo "" > "$clogf"
-	echo -e "- $MARKER rebase" | cat "$clogf.tmp" - >>
"$clogf"
-	rm "$clogf.tmp" "$clogf.unique"
+	# genlog.py always adds a Resolves: line, thus we
+	# can insert the rebase changelog item before it
+	sed -i "s/\(^Resolves:.*\)/- $MARKER rebase\n\1/" "$clogf"
 fi
 
 # during rh-dist-git genspec runs again and generates empty changelog

--
https://gitlab.com/cki-project/kernel-ark/-/merge_requests/1980

[OS-BUILD PATCHv4 5/8] redhat: move changelog addition to the spec file back into genspec.sh

[Herton R. Krzesinski (via Email Bridge)]

From: Herton R. Krzesinski <herton(a)redhat.com&gt;

redhat: move changelog addition to the spec file back into genspec.sh

In the next change, we want to detect at the Makefile level if we have new
changelog entries before we touch the spec file, so it makes sense to have
changelog addition being done in genspec.sh.

Signed-off-by: Herton R. Krzesinski <herton(a)redhat.com&gt;

diff --git a/redhat/genlog.sh b/redhat/genlog.sh
index blahblah..blahblah 100755
--- a/redhat/genlog.sh
+++ b/redhat/genlog.sh
@@ -36,13 +36,13 @@ git log --topo-order --no-merges -z "$GIT_NOTES"
"$GIT_FORMAT" \
 
 if [ "$HIDE_REDHAT" = "1" ]; then
 	grep -v -e "^- \[redhat\]" "$clogf" |
-		sed -e 's!\[Fedora\]!!g' > "$clogf.stripped"
-	cp "$clogf.stripped" "$clogf"
+		sed -e 's!\[Fedora\]!!g' > "$clogf.tmp"
+	mv -f "$clogf.tmp" "$clogf"
 fi
 
 if [ "$HIDE_UNSUPPORTED_ARCH" = "1" ]; then
-	grep -E -v "^-
\[(alpha|arc|arm|avr32|blackfin|c6x|cris|frv|h8300|hexagon|ia64|m32r|m68k|metag|microblaze|mips|mn10300|openrisc|parisc|score|sh|sparc|tile|um|unicore32|xtensa)\]"
"$clogf" > "$clogf.stripped"
-	cp "$clogf.stripped" "$clogf"
+	grep -E -v "^-
\[(alpha|arc|arm|avr32|blackfin|c6x|cris|frv|h8300|hexagon|ia64|m32r|m68k|metag|microblaze|mips|mn10300|openrisc|parisc|score|sh|sparc|tile|um|unicore32|xtensa)\]"
"$clogf" > "$clogf.tmp"
+	mv -f "$clogf.tmp" "$clogf"
 fi
 
 # If the markers aren't the same then this a rebase.
@@ -60,17 +60,4 @@ if [ "$LENGTH" = 0 ]; then
 	echo -n > "$clogf"
 fi
 
-cat "$clogf" "$SOURCES/$SPECCHANGELOG" > "$clogf.full"
-mv -f "$clogf.full" "$SOURCES/$SPECCHANGELOG"
-
-# genlog.py generates Resolves lines as well, strip these from RPM changelog
-grep -v -e "^Resolves: " "$SOURCES/$SPECCHANGELOG" >
"$clogf".stripped
-
-test -f "$SOURCES/$SPECFILE" &&
-	sed -i -e "
-	/%%SPECCHANGELOG%%/r $clogf.stripped
-	/%%SPECCHANGELOG%%/d" "$SOURCES/$SPECFILE"
-
 echo "MARKER is $MARKER"
-
-rm -f "$clogf".stripped;
diff --git a/redhat/genspec.sh b/redhat/genspec.sh
index blahblah..blahblah 100755
--- a/redhat/genspec.sh
+++ b/redhat/genspec.sh
@@ -88,9 +88,20 @@ if [ -n "$RHSELFTESTDATA" ]; then
 fi
 
 clogf=$(mktemp)
-trap 'rm -f "$clogf"' SIGHUP SIGINT SIGTERM EXIT
+trap 'rm -f "$clogf" "$clogf".stripped' SIGHUP SIGINT SIGTERM
EXIT
 "${0%/*}"/genlog.sh "$clogf"
 
+cat "$clogf" "$SOURCES/$SPECCHANGELOG" > "$clogf.full"
+mv -f "$clogf.full" "$SOURCES/$SPECCHANGELOG"
+
+# genlog.py generates Resolves lines as well, strip these from RPM changelog
+grep -v -e "^Resolves: " "$SOURCES/$SPECCHANGELOG" >
"$clogf".stripped
+
+test -f "$SOURCES/$SPECFILE" &&
+	sed -i -e "
+	/%%SPECCHANGELOG%%/r $clogf.stripped
+	/%%SPECCHANGELOG%%/d" "$SOURCES/$SPECFILE"
+
 if [ "$DISTRO" == "fedora" ]; then
 	# The tarball in the SRPM contains only the upstream sources.
 

--
https://gitlab.com/cki-project/kernel-ark/-/merge_requests/1980

[OS-BUILD PATCHv4 6/8] redhat: use new genlog.sh script to detect changes for dist-release

[Herton R. Krzesinski (via Email Bridge)]

From: Herton R. Krzesinski <herton(a)redhat.com&gt;

redhat: use new genlog.sh script to detect changes for dist-release

The current dist-release process has a flaw, where both its subtargets
(dist-release-changed and dist-release-finish) depends on setup-source
target to run (so genspec.sh is run twice). To avoid having to do the
work twice and do and revert changes before dist-release-finish is
called, we now detect changes by creating the changelog using genlog.sh
first to detect if there are really changes, and do the marker check
using an if condition instead of relying on git to see if there are
changes.

Signed-off-by: Herton R. Krzesinski <herton(a)redhat.com&gt;

diff --git a/redhat/Makefile b/redhat/Makefile
index blahblah..blahblah 100644
--- a/redhat/Makefile
+++ b/redhat/Makefile
@@ -572,18 +572,18 @@ dist-rpm-baseonly: dist-sources do-rpmbuild
 # unless you know what you're doing, you don't want to use the next four ones
 dist-release-finish: setup-source
 	@cp $(SOURCES)/$(SPECCHANGELOG) $(REDHAT)/$(SPECCHANGELOG)
+	@echo $(MARKER) > $(REDHAT)/marker
 	@$(GIT) add $(REDHAT)/$(SPECCHANGELOG)
 	@$(GIT) add $(REDHAT)/marker
 	@$(GIT) commit -s ../Makefile.rhelver $(REDHAT)/marker $(REDHAT)/$(SPECCHANGELOG)
$(SPECFILE).template -m "[redhat] $(RELEASETAG)"
-dist-release-changed: setup-source
-	@cp $(SOURCES)/$(SPECCHANGELOG) $(REDHAT)/$(SPECCHANGELOG)
-	@echo $(MARKER) > $(REDHAT)/marker
-	@# if neither changelog nor marker was updated, skip bumping a release
-	@$(GIT) update-index -q --really-refresh
-	@if $(GIT) diff-index --quiet HEAD; then \
+dist-release-changed:
+	@CLOGF="$$(mktemp)"; \
+	trap 'rm -f "$$CLOGF"' SIGHUP SIGINT SIGTERM EXIT; \
+	$(REDHAT)/genlog.sh "$$CLOGF"; \
+	if [ "$(MARKER)" == "$$(cat $(REDHAT)/marker)" ] && \
+	   [ "$$(wc -l $$CLOGF | cut -d ' ' -f 1)" -le 3 ]; then \
 		echo "Nothing changed, skipping updates"; \
 	else \
-		$(GIT) checkout -- $(REDHAT)/$(SPECCHANGELOG); \
 		$(REDHAT)/scripts/new_release.sh; \
 		$(MAKE) dist-release-finish; \
 	fi

--
https://gitlab.com/cki-project/kernel-ark/-/merge_requests/1980

[OS-BUILD PATCHv4 7/8] redhat: fix release number bump when dist-release-changed runs

[Herton R. Krzesinski (via Email Bridge)]

From: Herton R. Krzesinski <herton(a)redhat.com&gt;

redhat: fix release number bump when dist-release-changed runs

Right now inside the dist-release-changed target, we run a sub make
invocation calling the dist-release-finish target, after bumping
the release number with new_release.sh. The problem is that the
BUILD number still pointed to the old pre-bump number inside the
dist-release-finish invocation, and the problem is that we don't
reload the RHEL_RELEASE number in redhat/Makefile if BUILD is already
defined. Since it looks like that check was added to avoid overriding
the BUILD number set when RHSELFTESTDATA is defined, lets change
the if check there so BUILD has the updated value when dist-release-finish
runs through dist-release-changed.

Signed-off-by: Herton R. Krzesinski <herton(a)redhat.com&gt;

diff --git a/redhat/Makefile b/redhat/Makefile
index blahblah..blahblah 100644
--- a/redhat/Makefile
+++ b/redhat/Makefile
@@ -58,7 +58,7 @@ SPECKPATCHLEVEL:=$(shell $(GIT) show $(HEAD):Makefile | sed -ne
'/^PATCHLEVEL\ =
 SPECKSUBLEVEL:=$(shell $(GIT) show $(HEAD):Makefile | sed -ne '/^SUBLEVEL\ =\
/{s///;p;q}')
 SPECKEXTRAVERSION:=$(shell $(GIT) show $(HEAD):Makefile | sed -ne '/^EXTRAVERSION\ =\
/{s///;p;q}')
 GITID:= $(shell $(GIT) log --max-count=1 --pretty=format:%H $(HEAD))
-ifndef BUILD
+ifndef RHSELFTESTDATA
   BUILD:=$(RHEL_RELEASE)
 endif
 FLAVOR=

--
https://gitlab.com/cki-project/kernel-ark/-/merge_requests/1980

[OS-BUILD PATCHv4 8/8] redhat: fix dist-release build number test

[Herton R. Krzesinski (via Email Bridge)]

From: Herton R. Krzesinski <herton(a)redhat.com&gt;

redhat: fix dist-release build number test

Since previous commit, when fixing the build number bump, this test will
fail, now the build number and package release number should match as
this test was intended to check.

Signed-off-by: Herton R. Krzesinski <herton(a)redhat.com&gt;

diff --git a/redhat/self-test/2001-dist-release.bats
b/redhat/self-test/2001-dist-release.bats
index blahblah..blahblah 100644
--- a/redhat/self-test/2001-dist-release.bats
+++ b/redhat/self-test/2001-dist-release.bats
@@ -43,7 +43,6 @@ _dist-release_test_2() {
 	# Strip off ...kernel-VV.PP.SS-:
 	pkgrelease=${title##*kernel-+([5-9]).+([0-9]).+([0-9])-}
 	build=$(BUILD= DIST=.fc33 make dist-dump-variables | grep -E "^BUILD=" | cut
-d"=" -f2 | xargs)
-	((build--))
 	echo "pkgrelease=$pkgrelease"
 	echo "build=$build"
 	run _dist-release_test_2

--
https://gitlab.com/cki-project/kernel-ark/-/merge_requests/1980

[Herton R. Krzesinski (via Email Bridge)]

From: Herton R. Krzesinski on gitlab.com
https://gitlab.com/cki-project/kernel-ark/-/merge_requests/1980#note_1065...

Noticed a small error where marker would not updated, so did a new pushed with
it fixed.

[Jan Stancek (via Email Bridge)]

From: Jan Stancek on gitlab.com
https://gitlab.com/cki-project/kernel-ark/-/merge_requests/1980#note_1067...

Do I get it correctly that we still generate changelog twice?
```
_setup-source -> genspec.sh -> genlog.sh
dist-release-changed -> genlog.sh
```

[Herton R. Krzesinski (via Email Bridge)]

From: Herton R. Krzesinski on gitlab.com
https://gitlab.com/cki-project/kernel-ark/-/merge_requests/1980#note_1067...

Yes, we call genlog.sh twice, part of the check to see if there are changes
and if we need a release bump/commit. What is not called anymore twice is the
full genspec.sh/setup-source target.

[Jan Stancek (via Email Bridge)]

From: Jan Stancek on gitlab.com
https://gitlab.com/cki-project/kernel-ark/-/merge_requests/1980#note_1067...

Ok, thanks for confirming.

[Don Zickus (via Email Bridge)]

From: Don Zickus on gitlab.com
https://gitlab.com/cki-project/kernel-ark/-/merge_requests/1980#note_1067...

Thanks @hertonrk-rh for this series!  It looks to be an improvement in the
overall workflow (which shows with the later commits removing workarounds).  I
don't have any objections with the work.  It appears your solution is to
generate a temporary changelog and look for entries (wc -l > 3) and if so,
call dist-release-finish?

My only concern is, will there be confusion on genlog.sh vs. genlog.py?  I was
confused for awhile until I noticed the subtle '.sh' vs '.py'. Though it
looks
like genlog.py is only called by genlog.sh?  So perhaps it is fine.

[Herton R. Krzesinski (via Email Bridge)]

From: Herton R. Krzesinski on gitlab.com
https://gitlab.com/cki-project/kernel-ark/-/merge_requests/1980#note_1067...

Yes, the solution is to verify a temporary changelog instead of
changing/reverting/changing again using the Makefile target hacks: so
genlog.py always creates a header (the rpm changelog header for the changes
with the author and timestamp) plus a Resolves line, plus a blank line. Thus
that's 3 lines and why the check uses wc, so if there is only that, there are
no changes, and we use that to check for changes. Also due ark rebases there
is also the marker check.

I see genlog.sh vs .py might be confusing, but didn't have a better name.
Moving all code into python also may not be desirable, the python part does a
good job of parsing the git changelog (it replaced the previous awk script we
had), but the other stuff seems better to be handled through shell. And yes,
genlog.py is only called by genlog.sh.

[Don Zickus (via Email Bridge)]

From: Don Zickus on gitlab.com
https://gitlab.com/cki-project/kernel-ark/-/merge_requests/1980#note_1067...

Thanks!  I think this series is good prep work for the real fix, which is to
correctly handle the marker != last_marker case (merge update).  This patchset
will just say 'rebase', which is what the script comment suggests but isn't
accurate.  Instead this case happens with a nightly merge.

Do you have output of a simulated marker != last_marker changelog output? I
want Justin to verify that is what he expects it to look like.

[Herton R. Krzesinski (via Email Bridge)]

From: Herton R. Krzesinski on gitlab.com
https://gitlab.com/cki-project/kernel-ark/-/merge_requests/1980#note_1067...

I did a simulation here, using following steps, inside a kernel-ark checkout:

```
git checkout -b tmp b4f81e09279e3c37243c35a40947bab779eabd55
git merge split-loggen # this is my branch with changes of this MR
make LOCALVERSION="" dist-release
git merge origin/master
make LOCALVERSION="" dist-release
```

And this is the result of the last release commit with just the master merge:

```
$ git show -p f4a834d051e7cf13707b42475355cd79b79d2533
commit f4a834d051e7cf13707b42475355cd79b79d2533 (HEAD -> tmp)
Author: Herton R. Krzesinski <herton(a)redhat.com&gt;
Date:   Wed Aug 17 12:50:22 2022 -0300

    [redhat] kernel-6.0.0-0.rc1.3cc40a443a04.15

    Signed-off-by: Herton R. Krzesinski <herton(a)redhat.com&gt;

diff --git a/Makefile.rhelver b/Makefile.rhelver
index 5e65c2fc923e..39d4dc1fe57d 100644
--- a/Makefile.rhelver
+++ b/Makefile.rhelver
@@ -12,7 +12,7 @@ RHEL_MINOR = 99
 #
 # Use this spot to avoid future merge conflicts.
 # Do not trim this comment.
-RHEL_RELEASE = 14
+RHEL_RELEASE = 15

 #
 # ZSTREAM
diff --git a/redhat/kernel.changelog-9.99 b/redhat/kernel.changelog-9.99
index 8d861210ed9e..337912c32c5f 100644
--- a/redhat/kernel.changelog-9.99
+++ b/redhat/kernel.changelog-9.99
@@ -1,3 +1,7 @@
+* Wed Aug 17 2022 Herton R. Krzesinski <herton(a)redhat.com&gt;
[6.0.0-0.rc1.3cc40a443a04.15]
+- 3cc40a443a04 rebase
+Resolves:
+
 * Wed Aug 17 2022 Herton R. Krzesinski <herton(a)redhat.com&gt; [6.0.0-0.rc1.14]
 - redhat: fix dist-release build number test (Herton R. Krzesinski)
 - redhat: fix release number bump when dist-release-changed runs (Herton R.
Krzesinski)
diff --git a/redhat/marker b/redhat/marker
index 60cf3a175526..6fa36a12cba0 100644
--- a/redhat/marker
+++ b/redhat/marker
@@ -1 +1 @@
-v6.0-rc1
+3cc40a443a04
```

[Don Zickus (via Email Bridge)]

From: Don Zickus on gitlab.com
https://gitlab.com/cki-project/kernel-ark/-/merge_requests/1980#note_1067...

Perfect example env!  So a couple of things.  I believe the 'rebase' word is
incorrect and the 'Resolves' line should be removed?? (I thought it was with
your sed magic?).

Sadly, I think we should fix these be committing this MR otherwise we will
start seeing lots of 'rebase' in the changelog every night, which isn't a
good
idea.

@jmflinuxtx - would is your expected changelog output for nightly merges?

"Merged upstream commit 3cc40a443a04" ??

[Justin M. Forbes (via Email Bridge)]

From: Justin M. Forbes on gitlab.com
https://gitlab.com/cki-project/kernel-ark/-/merge_requests/1980#note_1067...

I actually prefer something along the lines of:

`- Linux v6.0.0-0.rc1.3cc40a443a04`

It technically isn't a rebase, it is a merge, but for as long as I can recall,
prior to ark, Fedora always had a changelog entry  of "Linux
v<upstream-ver>"

[Herton R. Krzesinski (via Email Bridge)]

From: Herton R. Krzesinski on gitlab.com
https://gitlab.com/cki-project/kernel-ark/-/merge_requests/1980#note_1067...

Ok I'll rework the changelog entry. About the Resolves line, it's kept only in
the kernel.changelog file, the rpm/dist-git should have it removed. We need to
keep the Resolves line in the redhat/kernel.changelog* since it's later used.