sysvinit-3.05

[ken@…]

As reported by Uwe Duffert in -dev, there is a new location for sysvinit, ​https://github.com/slicer69/sysvinit/releases/

From the closed issues there, this does seem to be the new home for what was at savannah.

Therefore, we ought to consider moving to this, and fixing the php script.

However, at the moment the new translated manpages appear to be in non-standard locations and using ROOT= (for DESTDIR) fails on the first (man1 de) pages because the directories are not created. ​https://github.com/slicer69/sysvinit/issues/14

If these two points are satisfactorily resolved, maybe we should provide a patch for building on LFS to save carrying our own patch to remove certain items ?

[Bruce Dubbs]

Our patch to remove programs superseded by other packages works without changes.

This packages does not use configure, so a change to remove wall, mountpoint, last, lastb, mesg, sulogin, and utmpdump can't be done with --without-<program>

There are three Makefiles: ./man/Makefile, ./Makefile, and ./src/Makefile.

In ./Makefile the only relevant portion is:

all install clean distclean:
   @rm -f $(PACKAGE)-$(VERSION).tar.xz $(PACKAGE)-$(VERSION).tar.xz.sig
   $(MAKE) VERSION=$(VERSION) -C man $@
   $(MAKE) VERSION=$(VERSION) -C src $@

In man/Makefile there is no install. It is all in src/Makefile.

In src/Makefile we might be able to create a different patch like:

ifeq ($(DISTRO),Debian)
CPPFLAGS+= -DACCTON_OFF
SBIN  += sulogin bootlogd
MAN8  += sulogin.8 bootlogd.8
MANDB :=
endif

instead of our current patch.

For make install we can s/ROOT/DESTDIR/g

It wants to install in /bin, /sbin, and (for wall and utmpdump only), /usr/bin. We can let our symlinks for /bin and /sbin take care of that.

In the install section for manpages is:

for lang in  '' $(subst ../man/,,$(wildcard ../man/po/??/)); do \
         $(INSTALL_DIR) $(ROOT)$(MANDIR)/man1/$$lang; \
         $(INSTALL_DIR) $(ROOT)$(MANDIR)/man5/$$lang; \
         $(INSTALL_DIR) $(ROOT)$(MANDIR)/man8/$$lang; \
      done

Those should be .e.g. '$(INSTALL_DIR) $(ROOT)$(MANDIR)/$$lang/man1;'

For the actual install they are doing:

      for man in $(MAN1) $(subst ../man/,,$(foreach man,$(MAN1),$(wildcard ../man/??/$(man)))); do \
         $(INSTALL_DATA) ../man/$$man $(ROOT)$(MANDIR)/man1/$$man; \
         sed -i "1{ $(MANDB); }" $(ROOT)$(MANDIR)/man1/$$man ; \
      done

I'm not really following this so I'm not sure how to change this for the translated man pages. Perhaps all that is needed is to change in the for statement

$(wildcard ../man/??/$(man))

to $(wildcard ../??/man/$(man))

[ken@…]

I've got problems with my hub at the moment (red LED which suggests overheating) and I've got an online course and phone calls on Monday and Tuesday (phone connected via hub) so my online time is going to be limited.

I've noted the suggestions, might try those while I'm offline.

[Bruce Dubbs]

I've been trying to fix this today. I've found the problem in src/Makefile, but I can't figure out a solution.

for man in $(MAN5) $(subst ../man/,,$(foreach man,$(MAN5),$(wildcard ../man/??/$(man)))); do

assigns the variable man to, for instance:

man = initscript.5
man = de/initscript.5
...

It then wants to install as

$(INSTALL_DATA) ../man/$$man $(ROOT)$(MANDIR)/man5/$$man

but it needs to separate PO from FILENAME and use:

$(INSTALL_DATA) ../man/$$man $(ROOT)$(MANDIR)/$(PO)/man5/$(FILENAME)

It's ok if PO is null. The problem is that make does not like assigning variables inside a loop. If someone has a solution, I'd like to see it.

[ken@…]

Thanks for that. I tried your earlier suggestion for the translations with ROOT=, that installed apparently successfully except that no translations were installed!

install -m 644 initreq.h /tmp/SYSV99/usr/include/
for lang in  '' ; do \
		install -m 755 -d /tmp/SYSV99/usr/share/man/$lang/man1; \
		install -m 755 -d /tmp/SYSV99/usr/share/man/$lang/man5; \
		install -m 755 -d /tmp/SYSV99/usr/share/man/$lang/man8; \
	done
for man in last.1 lastb.1 mesg.1 readbootlog.1 utmpdump.1 wall.1      ; do \
		install -m 644 ../man/$man /tmp/SYSV99/usr/share/man/man1/$man; \
		sed -i "1{ s@^\('\\\\\"\)[^\*-]*-\*- coding: [^[:blank:]]\+ -\*-@\1@; }" /tmp/SYSV99/usr/share/man/man1/$man ; \
	done
for man in last.1 lastb.1 mesg.1 readbootlog.1 utmpdump.1 wall.1      ; do \
		install -m 644 ../man/$man /tmp/SYSV99/usr/share/man/man5/$man; \
		sed -i "1{ s@^\('\\\\\"\)[^\*-]*-\*- coding: [^[:blank:]]\+ -\*-@\1@; }" /tmp/SYSV99/usr/share/man/man5/$man ; \
	done
for man in halt.8 init.8 killall5.8 pidof.8 poweroff.8 reboot.8 runlevel.8 shutdown.8 telinit.8 fstab-decode.8 logsave.8 sulogin.8 bootlogd.8      ; do \
		install -m 644 ../man/$man /tmp/SYSV99/usr/share/man/man8/$man; \
		sed -i "1{ s@^\('\\\\\"\)[^\*-]*-\*- coding: [^[:blank:]]\+ -\*-@\1@; }" /tmp/SYSV99/usr/share/man/man8/$man ; \
	done
make[1]: Leaving directory '/tmp/sysvinit-3.05/src'

I then went back to vanilla 3.0.5 with ROOT= and noted:

install -m 644 initreq.h /tmp/SYSVDEF/usr/include/
for lang in  '' ; do \
		install -m 755 -d /tmp/SYSVDEF/usr/share/man/man1/$lang; \
		install -m 755 -d /tmp/SYSVDEF/usr/share/man/man5/$lang; \
		install -m 755 -d /tmp/SYSVDEF/usr/share/man/man8/$lang; \
	done

i.e. at that point $lang is not set by upstream leading to

install: cannot create regular file '/tmp/SYSVDEF/usr/share/man/man1/de/last.1': No such file or directory
sed: can't read /tmp/SYSVDEF/usr/share/man/man1/de/last.1: No such file or directory

Since this is only an "add translations" release I'm inclined to wait a while to see if upstream responds to the issue I raised.

[Bruce Dubbs]

I've found a solution by rewriting parts of the Makefile. If it works for us, we can submit it upstream. I'll try to get a new patch out late tomorrow.

One issue for LFS is whether we should continue to install bootlogd and fstab-decode. I don't see any use for those packages in LFS.

[Bruce Dubbs]

I'd like to get someone to try the attached patch. The build is slightly more complex than what we have now:

The patch only changes src/Makefile. There are three GCC warnings on the packages we build, but they appear to be harmless.

patch -Np1 -i ../sysvinit-3.05-consolidated-1.patch &&
make DISTRO=LFS &&
make DESTDIR=$PWD/install DISTRO=LFS install 

Of course a direct install can be made at root without DESTDIR. Build time at -j1 is 1.5 seconds on my system.

[ken@…]

Replying to Bruce Dubbs:

I'd like to get someone to try the attached patch. The build is slightly more complex than what we have now:

The patch only changes src/Makefile. There are three GCC warnings on the packages we build, but they appear to be harmless.

patch -Np1 -i ../sysvinit-3.05-consolidated-1.patch &&
make DISTRO=LFS &&
make DESTDIR=$PWD/install DISTRO=LFS install 

Of course a direct install can be made at root without DESTDIR. Build time at -j1 is 1.5 seconds on my system.

Works for me, thanks.

[thomas]

Replying to Bruce Dubbs:

I'd like to get someone to try the attached patch. The build is slightly more complex than what we have now:

The patch only changes src/Makefile. There are three GCC warnings on the packages we build, but they appear to be harmless.

patch -Np1 -i ../sysvinit-3.05-consolidated-1.patch &&
make DISTRO=LFS &&
make DESTDIR=$PWD/install DISTRO=LFS install 

Of course a direct install can be made at root without DESTDIR. Build time at -j1 is 1.5 seconds on my system.

i just executed the three cmds as shown above. In usr/bin i got:

thomas@lfs-x86_64:~/sysvinit-3.05/install$ dir usr/bin/
insgesamt 8
drwxr-xr-x 2 thomas users 4096  5. Nov 20:59 .
drwxr-xr-x 5 thomas users 4096  5. Nov 20:59 ..
lrwxrwxrwx 1 thomas users    4  5. Nov 20:59 lastb -> last
thomas@lfs-x86_64:~/sysvinit-3.05/install$ 

the lastb points to nowhere as last does not exist there.

[Bruce Dubbs]

Thanks Thomas. I thought I had fixed that, but it didn't make it into the patch. I'll upload a new patch in a few minutes.

[Bruce Dubbs]

OK, try with sysvinit-3.05-consolidated-2.patch

[thomas]

I'd suggest two small fixes:

$ diff -Naur sysvinit-3.05-consolidated-2.patch{.orig,}
--- sysvinit-3.05-consolidated-2.patch.orig	2022-11-06 08:20:30.905254677 +0100
+++ sysvinit-3.05-consolidated-2.patch	2022-11-06 08:21:11.995763315 +0100
@@ -11,7 +11,7 @@
 +MAN1  =
 +MAN5  = initscript.5 inittab.5 initctl.5
 +MAN8  = halt.8 init.8 killall5.8 pidof.8 poweroff.8 reboot.8 runlevel.8
-+MAN8  = shutdown.8 telinit.8
++MAN8  += shutdown.8 telinit.8
 +MANDB:=
 +endif
 +
@@ -113,7 +113,7 @@
 +			done; \
 +			for man in $(MAN8); do \
 +				if ! test -f "../man/$$lang/$$man"; then continue; fi; \
-+				$(INSTALL_DATA) ../man/$$lang/$$man $(DESTDIR)/usr/share/man/$$lang/man5/$$man; \
++				$(INSTALL_DATA) ../man/$$lang/$$man $(DESTDIR)/usr/share/man/$$lang/man8/$$man; \
 +			done; \
  		done
 -ifeq ($(ROOT),)

[Bruce Dubbs]

You're right. Both of those are what I intended. It really helps to have someone else review the patch.

I've pushed the fixed patch to the patches git repository as sysvinit-3.05-consolidated-1.patch. I'll update LFS to the new sysvinit with this patch at the usual mid month update.

[Xi Ruoyao]

Fixed at r11.2-173-g9292384ba.