Skip to contentSkip to conversation

package/ecryptfs-utils: fix parallel build

PATCH Superseded Fabrice Fontaine 2 participants 4 messages Dec 26, 2019

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 package/ecryptfs-utils: fix parallel build

Build randomly fails since December 2017 on:

make[5]: Entering directory '/home/buildroot/autobuild/instance-2/output-1/build/ecryptfs-utils-111/src/utils'
/bin/mkdir -p '/home/buildroot/autobuild/instance-2/output-1/target/sbin'
/bin/bash ../../libtool --mode=install /usr/bin/install -c mount.ecryptfs umount.ecryptfs mount.ecryptfs_private '/home/buildroot/autobuild/instance-2/output-1/target/sbin'
libtool: install: /usr/bin/install -c mount.ecryptfs /home/buildroot/autobuild/instance-2/output-1/target/sbin/mount.ecryptfs /usr/bin/install: cannot create regular file '/home/buildroot/autobuild/instance-2/output-1/target/sbin/mount.ecryptfs': File exists Makefile:832: recipe for target 'install-rootsbinPROGRAMS' failed
make[5]: *** [install-rootsbinPROGRAMS] Error 1

Fixes:
- http://autobuild.buildroot.org/results/2063bd18830ef02dca0bed889c43639009bc688e

---
Changes v1 -> v2 (after review of Thomas Petazzoni):
- Fix parallel build issue instead of disabling it

 .../0003-fix-parallel-build-issue.patch       | 60 +++++++++++++++++++
 .../0004-drop-AM_GLIB_GNU_GETTEXT.patch       | 23 +++++++
 package/ecryptfs-utils/ecryptfs-utils.mk      |  2 +
 3 files changed, 85 insertions(+)
 create mode 100644 package/ecryptfs-utils/0003-fix-parallel-build-issue.patch
 create mode 100644 package/ecryptfs-utils/0004-drop-AM_GLIB_GNU_GETTEXT.patch
package/ecryptfs-utils/0003-fix-parallel-build-issue.patch new file +60 -0
new file mode 100644
index 0000000000..35b245c94f
@@ -0,0 +1,60 @@
+fix parallel build issue
+
+Build randomly fails since December 2017 on buildroot
+(http://autobuild.buildroot.org/?reason=ecryptfs-utils-111):
+
+make[5]: Entering directory '/home/buildroot/autobuild/instance-2/output-1/build/ecryptfs-utils-111/src/utils'
+  /bin/mkdir -p '/home/buildroot/autobuild/instance-2/output-1/target/sbin'
+  /bin/bash ../../libtool   --mode=install /usr/bin/install -c mount.ecryptfs umount.ecryptfs mount.ecryptfs_private '/home/buildroot/autobuild/instance-2/output-1/target/sbin'
+libtool: install: /usr/bin/install -c mount.ecryptfs /home/buildroot/autobuild/instance-2/output-1/target/sbin/mount.ecryptfs
+/usr/bin/install: cannot create regular file '/home/buildroot/autobuild/instance-2/output-1/target/sbin/mount.ecryptfs': File exists
+Makefile:832: recipe for target 'install-rootsbinPROGRAMS' failed
+make[5]: *** [install-rootsbinPROGRAMS] Error 1
+
+As spotted by Thomas Petazzoni, build failure happens because of the
+following line in src/utils/Makefile.am: 
+
+install-exec-hook:      install-rootsbinPROGRAMS
+        -rm -f "$(DESTDIR)/$(rootsbindir)/umount.ecryptfs_private"
+        $(LN_S) "mount.ecryptfs_private" "$(DESTDIR)/$(rootsbindir)/umount.ecryptfs_private"
+
+The install-exec-hook target should not have a dependency on
+install-rootsbinPROGRAMS.
+
+From https://www.gnu.org/software/automake/manual/html_node/Extending.html#Extending:
+
+"""
+In contrast, some rules also have a way to run another rule, called a
+hook; hooks are always executed after the main rule?s work is done. The
+hook is named after the principal target, with ?-hook? appended. The
+targets allowing hooks are install-data, install-exec, uninstall, dist,
+and distcheck.
+
+For instance, here is how to create a hard link to an installed program:
+
+install-exec-hook:
+        ln $(DESTDIR)$(bindir)/program$(EXEEXT) \
+           $(DESTDIR)$(bindir)/proglink$(EXEEXT)
+
+"""
+
+So, they explicitly say that these hooks are run after the main rule
+work is done, which means the dependency on install-rootsbinPROGRAMS is
+not needed. And the example they use to illustrate is *exactly* the
+situation of ecryptfs-utils: creating a link to a program that was
+installed.
+
+Signed-off-by: Fabrice Fontaine <[email protected]>
+[Upstream status: https://bugs.launchpad.net/ecryptfs/+bug/1857622]
+
+diff -Nuar ecryptfs-utils-111-orig/src/utils/Makefile.am ecryptfs-utils-111/src/utils/Makefile.am
+--- ecryptfs-utils-111-orig/src/utils/Makefile.am	2019-12-26 15:14:16.656146065 +0100
++++ ecryptfs-utils-111/src/utils/Makefile.am	2019-12-26 15:14:30.160146621 +0100
+@@ -67,6 +67,6 @@
+ test_SOURCES = test.c io.c
+ test_LDADD = $(top_builddir)/src/libecryptfs/libecryptfs.la
+ 
+-install-exec-hook:	install-rootsbinPROGRAMS
++install-exec-hook:
+ 	-rm -f "$(DESTDIR)/$(rootsbindir)/umount.ecryptfs_private"
+ 	$(LN_S) "mount.ecryptfs_private" "$(DESTDIR)/$(rootsbindir)/umount.ecryptfs_private"
package/ecryptfs-utils/0004-drop-AM_GLIB_GNU_GETTEXT.patch new file +23 -0
new file mode 100644
index 0000000000..0c8d0ec36e
@@ -0,0 +1,23 @@
+drop AM_GLIB_GNU_GETTEXT
+
+This will fix the following failure when running autoreconf:
+
+configure.ac:372: warning: macro 'AM_GLIB_GNU_GETTEXT' not found in library
+configure.ac:372: error: possibly undefined macro: AM_GLIB_GNU_GETTEXT
+      If this token and others are legitimate, please use m4_pattern_allow.
+      See the Autoconf documentation.
+
+Signed-off-by: Fabrice Fontaine <[email protected]>
+
+diff -Nuar ecryptfs-utils-111-orig/configure.ac ecryptfs-utils-111/configure.ac
+--- ecryptfs-utils-111-orig/configure.ac	2019-12-26 15:14:16.652146065 +0100
++++ ecryptfs-utils-111/configure.ac	2019-12-26 15:54:49.212246135 +0100
+@@ -369,8 +369,6 @@
+ AC_DEFINE_UNQUOTED(GETTEXT_PACKAGE,"$GETTEXT_PACKAGE",
+                    [the gettext translation domain])
+ 
+-AM_GLIB_GNU_GETTEXT
+-
+ IT_PROG_INTLTOOL([0.41.0])
+ 
+ LOCALEDIR='${prefix}/${datadir}/locale'
package/ecryptfs-utils/ecryptfs-utils.mk +2 -0
index 326ff4f2dd..78fe2fb499 100644
@@ -9,6 +9,8 @@ ECRYPTFS_UTILS_SOURCE = ecryptfs-utils_$(ECRYPTFS_UTILS_VERSION).orig.tar.gz
 ECRYPTFS_UTILS_SITE = https://launchpad.net/ecryptfs/trunk/$(ECRYPTFS_UTILS_VERSION)/+download
 ECRYPTFS_UTILS_LICENSE = GPL-2.0+
 ECRYPTFS_UTILS_LICENSE_FILES = COPYING
+# We're patching src/utils/Makefile.am
+ECRYPTFS_UTILS_AUTORECONF = YES
 
 ECRYPTFS_UTILS_DEPENDENCIES = keyutils libnss host-intltool
 ECRYPTFS_UTILS_CONF_OPTS = --disable-pywrap
Signature · 1 line
2.24.0
#1 lore raw

PATCH v2 1/1 package/ecryptfs-utils: fix parallel build

Hello Fabrice,

Thanks for the new iteration!

Quoted 15 lines · On Thu, 26 Dec 2019 16:03:30 +0100 Fabrice Fontaine…
package/ecryptfs-utils/0004-drop-AM_GLIB_GNU_GETTEXT.patch new file +9 -0
new file mode 100644
index 0000000000..0c8d0ec36e
@@ -0,0 +1,23 @@
+drop AM_GLIB_GNU_GETTEXT
+
+This will fix the following failure when running autoreconf:
+
+configure.ac:372: warning: macro 'AM_GLIB_GNU_GETTEXT' not found in library
+configure.ac:372: error: possibly undefined macro: AM_GLIB_GNU_GETTEXT
+      If this token and others are legitimate, please use m4_pattern_allow.
+      See the Autoconf documentation.
+

package/ecryptfs-utils/0004-drop-AM_GLIB_GNU_GETTEXT.patch line 19

Quoted 10 lines
package/ecryptfs-utils/0004-drop-AM_GLIB_GNU_GETTEXT.patch new file +10 -0
+Signed-off-by: Fabrice Fontaine <[email protected]>
+
+diff -Nuar ecryptfs-utils-111-orig/configure.ac ecryptfs-utils-111/configure.ac
+--- ecryptfs-utils-111-orig/configure.ac	2019-12-26 15:14:16.652146065 +0100
++++ ecryptfs-utils-111/configure.ac	2019-12-26 15:54:49.212246135 +0100
+@@ -369,8 +369,6 @@
+ AC_DEFINE_UNQUOTED(GETTEXT_PACKAGE,"$GETTEXT_PACKAGE",
+                    [the gettext translation domain])
+ 
+-AM_GLIB_GNU_GETTEXT

Are you sure it is useless? Just dropping a line of code that is "annoying" because it causes a build failure is quite weird.

What makes you think it is useless? This m4 macro is most likely provided by some other package.

Thomas

Signature · 3 lines
Thomas Petazzoni, CTO, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com
#2 lore raw

PATCH v2 1/1 package/ecryptfs-utils: fix parallel build

Hi Thomas,

Le jeu. 26 d?c. 2019 ? 16:10, Thomas Petazzoni
<[email protected]> a ?crit :

Quoted 37 lines

Hello Fabrice,

Thanks for the new iteration!

Quoted 15 lines · On Thu, 26 Dec 2019 16:03:30 +0100 Fabrice Fontaine…
package/ecryptfs-utils/0004-drop-AM_GLIB_GNU_GETTEXT.patch new file +9 -0
new file mode 100644
index 0000000000..0c8d0ec36e
@@ -0,0 +1,23 @@
+drop AM_GLIB_GNU_GETTEXT
+
+This will fix the following failure when running autoreconf:
+
+configure.ac:372: warning: macro 'AM_GLIB_GNU_GETTEXT' not found in library
+configure.ac:372: error: possibly undefined macro: AM_GLIB_GNU_GETTEXT
+      If this token and others are legitimate, please use m4_pattern_allow.
+      See the Autoconf documentation.
+

package/ecryptfs-utils/0004-drop-AM_GLIB_GNU_GETTEXT.patch line 19

Quoted 10 lines
package/ecryptfs-utils/0004-drop-AM_GLIB_GNU_GETTEXT.patch new file +10 -0
+Signed-off-by: Fabrice Fontaine <[email protected]>
+
+diff -Nuar ecryptfs-utils-111-orig/configure.ac ecryptfs-utils-111/configure.ac
+--- ecryptfs-utils-111-orig/configure.ac     2019-12-26 15:14:16.652146065 +0100
++++ ecryptfs-utils-111/configure.ac  2019-12-26 15:54:49.212246135 +0100
+@@ -369,8 +369,6 @@
+ AC_DEFINE_UNQUOTED(GETTEXT_PACKAGE,"$GETTEXT_PACKAGE",
+                    [the gettext translation domain])
+
+-AM_GLIB_GNU_GETTEXT

Are you sure it is useless? Just dropping a line of code that is "annoying" because it causes a build failure is quite weird.

Quoted 2 lines

What makes you think it is useless? This m4 macro is most likely provided by some other package.

Yes, it is provided by libglib2 but we definitely does not want to depend on it. I saw that this issue was already raised by avahi in 2014:
http://lists.busybox.net/pipermail/buildroot/2014-November/111185.html.
After reading the 4 proposals from this thread, I decided to go to the third one as the package already has a dependency on host-intltool. But frankly, upstream seems pretty much dead (no updates since June 2017) so my preferred option would be to just disable parallel build.

Quoted 6 lines

Thomas
--
Thomas Petazzoni, CTO, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com

Best Regards,

Fabrice

#3 lore raw

PATCH v2 1/1 package/ecryptfs-utils: fix parallel build

Quoted 7 lines · On Thu, 26 Dec 2019 16:34:19 +0100 Fabrice Fontaine…

Yes, it is provided by libglib2 but we definitely does not want to depend on it. I saw that this issue was already raised by avahi in 2014:
http://lists.busybox.net/pipermail/buildroot/2014-November/111185.html.
After reading the 4 proposals from this thread, I decided to go to the third one as the package already has a dependency on host-intltool. But frankly, upstream seems pretty much dead (no updates since June 2017) so my preferred option would be to just disable parallel build.

Then perhaps we should make an exception and patch the Makefile.in directly?

Thomas

Signature · 3 lines
Thomas Petazzoni, CTO, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com