NetBSD Problem Report #38571
From apb@cequrux.com Sat May 3 21:15:52 2008
Return-Path: <apb@cequrux.com>
Received: from mail.netbsd.org (mail.netbsd.org [204.152.190.11])
by narn.NetBSD.org (Postfix) with ESMTP id D319763B8BC
for <gnats-bugs@gnats.NetBSD.org>; Sat, 3 May 2008 21:15:52 +0000 (UTC)
Message-Id: <20080503211537.379E28A496D@apb-laptoy.apb.alt.za>
Date: Sat, 3 May 2008 23:15:37 +0200 (SAST)
From: apb@cequrux.com
Reply-To: apb@cequrux.com
To: gnats-bugs@gnats.NetBSD.org
Subject: sysinst runs "postinstall fix"
X-Send-Pr-Version: 3.95
>Number: 38571
>Category: install
>Synopsis: sysinst runs "postinstall fix"
>Confidential: no
>Severity: serious
>Priority: high
>Responsible: install-manager
>State: open
>Class: sw-bug
>Submitter-Id: net
>Arrival-Date: Sat May 03 21:20:00 +0000 2008
>Last-Modified: Sun Sep 16 01:05:02 +0000 2012
>Originator: Alan Barrett
>Release: NetBSD 4.99.60
>Organization:
Not much
>Environment:
System: NetBSD 4.99.60 i386
>Description:
If you use sysinst to upgrade to a new vesion of NetBSD, then sysinst
will "/usr/sbin/postinstall -s /.sysinst -d / fix". This will overwrite
any local changes to files in several parts of /etc, leading to data
loss if there are inadequate backups.
>How-To-Repeat:
Code inspection
>Fix:
Run "postinstall check" instead of "postinstall fix"; display the results,
and obtain permission from the user before running "postinstall fix".
Alternatively, offer to run "etcupdate" instead of "postinstall fix".
>Audit-Trail:
From: Julian Djamil Fagir <gnrp@komkon2.de>
To: gnats-bugs@gnats.netbsd.org
Cc:
Subject: Re: install/38571: sysinst runs "postinstall fix"
Date: Wed, 29 Feb 2012 02:30:29 +0100
--MP_/joB4JmgFVA+c7H_120kwHN4
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Content-Disposition: inline
Hi,
the attached patch fixes that issue for me. On the one hand, it introduces a
new function that exits curses, executes a command and afterwards returns.
This function is used to execute etcupdate(8) in case the user wants it.
Please carefully look at that patch, I only have one (virtual) machine to
test sysinst, and I'm not that sure about the curses part (though it works
for me).
Regards, Julian
PS: In advance: I'm sorry if my mailer breaks encoding again by making it
quoted-printable.
--MP_/joB4JmgFVA+c7H_120kwHN4
Content-Type: text/x-patch
Content-Transfer-Encoding: 7bit
Content-Disposition: attachment; filename=defs.h.diff
--- defs.h
+++ defs.h
@@ -400,10 +400,11 @@
void mnt_net_config(void);
/* From run.c */
int collect(int, char **, const char *, ...) __printflike(3, 4);
int run_program(int, const char *, ...) __printflike(2, 3);
+int launch_command(int, const char *);
void do_logging(void);
int do_system(const char *);
/* from upgrade.c */
void do_upgrade(void);
--MP_/joB4JmgFVA+c7H_120kwHN4
Content-Type: text/x-patch
Content-Transfer-Encoding: 7bit
Content-Disposition: attachment; filename=menus.mi.diff
--- menus.mi
+++ menus.mi
@@ -374,10 +374,16 @@
yesno = 0;
#endif
};
option MSG_other, exit, action
{ yesno = 0; };
+
+menu runpostinst, title MSG_Run_Postinst, y=-5;
+ option MSG_Postinst, exit, action { *(int *)arg = 0; };
+ option MSG_Etcupdate, exit, action { *(int *)arg = 1; };
+ option MSG_Skip_postinst, exit, action { *(int *)arg = 2; };
+ option MSG_Abandon, exit, action { *(int *)arg = 3; };
menu rootsh, title MSG_Root_shell;
option "/bin/sh", exit, action {*(const char **)arg = "/bin/sh";};
option "/bin/ksh", exit, action {*(const char **)arg = "/bin/ksh";};
option "/bin/csh", exit, action {*(const char **)arg = "/bin/csh";};
--MP_/joB4JmgFVA+c7H_120kwHN4
Content-Type: text/x-patch
Content-Transfer-Encoding: quoted-printable
Content-Disposition: attachment; filename=msg.mi.de.diff
--- msg.mi.de
+++ msg.mi.de
@@ -106,10 +106,22 @@
Zu diesem Zeitpunkt sollten Sie bereits eine vollst=E4ndige
Datensicherung durchgef=FChrt haben! Wollen Sie NetBSD wirklich aktualisie=
ren?
(Dies ist die letzte Warnung, bevor Ihre Festplatte(n) modifiziert werden.)
}
=20
+message Run_Postinst
+{Dateien in /etc updaten}
+
+message Postinst
+{postinst ausf=FChren (existierende Konfiguration =FCberschreiben)}
+
+message Etcupdate
+{etcupdate ausf=FChren (zu jeder ge=E4nderten Datei nachfragen)}
+
+message Skip_postinst
+{Keine Dateien in /etc updaten}
+
message reinstallusure
{Im folgenden werden die NetBSD Distributionssets (Kernel + Basissystem)
auf eine vorbereitete Festplatte ausgepackt. Diese Prozedur l=E4dt und ent=
packt
die Sets auf eine im Vorfeld partitionierte und bootf=E4hige Festplatte au=
s.
Es werden weder Festplatten gelabelt, Bootbl=F6cke aktualisiert noch beste=
hende
@@ -122,10 +134,13 @@
Wollen Sie die NetBSD Distributionssets wirklich erneut installieren?
(Dies ist die letzte Warnung, bevor die Inhalte Ihres Dateisystems
=FCberschrieben werden!)
}
=20
+message mount_failed
+{Versuch, %s zu mounten ist fehlgeschlagen. Fortfahren?
+}
=20
message nodisk
{Ich kann keine f=FCr NetBSD nutzbaren Festplatten finden.
Zur=FCck zum Hauptmen=FC...
}
@@ -709,13 +724,13 @@
{Erstelle Ger=E4tedateien in /dev ...
}
=20
message badfs
{Das Dateisystem auf /dev/%s%c scheint kein BSD-Dateisystem zu sein,
-die Pr=FCfung des Dateisystems (fsck) ist fehlgeschlagen.
+die Pr=FCfung des Dateisystems (fsck) ist fehlgeschlagen (Fehler %d).
=20
-Die Aktualisierung wird abgebrochen. (Fehlernummer %d.)
+Die Aktualisierung trotzdem fortsetzen?
}
=20
message rootmissing
{Das Zielverzeichnis %s existiert nicht.
}
--MP_/joB4JmgFVA+c7H_120kwHN4
Content-Type: text/x-patch
Content-Transfer-Encoding: 7bit
Content-Disposition: attachment; filename=msg.mi.en.diff
--- msg.mi.en
+++ msg.mi.en
@@ -102,10 +102,22 @@
before this procedure! Do you really want to upgrade NetBSD?
(This is your last warning before this procedure starts modifying your
disks.)
}
+message Run_Postinst
+{Update files in /etc}
+
+message Postinst
+{Execute postinst (overwrite existing configuration)}
+
+message Etcupdate
+{Execute etcupdate (ask for every overwrite)}
+
+message Skip_postinst
+{Don't update any file on /etc}
+
message reinstallusure
{Ok, let's unpack the NetBSD distribution sets to a bootable hard disk.
This procedure just fetches and unpacks sets onto an pre-partitioned
bootable disk. It does not label disks, upgrade bootblocks, or save
any existing configuration info. (Quit and choose `install' or
@@ -115,10 +127,13 @@
Do you really want to reinstall NetBSD distribution sets?
(This is your last warning before this procedure starts modifying your
disks.)
}
+message mount_failed
+{Mounting %s failed. Continue?
+}
message nodisk
{I can not find any hard disks for use by NetBSD. You will be
returned to the original menu.
}
@@ -678,11 +693,11 @@
{Making device nodes ...
}
message badfs
{It appears that /dev/%s%c is not a BSD file system or the fsck was
-not successful. The upgrade has been aborted. (Error number %d.)
+not successful (Error number %d.). Try mounting it anyway?
}
message rootmissing
{ target root is missing %s.
}
--MP_/joB4JmgFVA+c7H_120kwHN4
Content-Type: text/x-patch
Content-Transfer-Encoding: quoted-printable
Content-Disposition: attachment; filename=msg.mi.es.diff
--- msg.mi.es
+++ msg.mi.es
@@ -108,10 +108,22 @@
completa antes de este procedimiento! =BFRealmente desea actualizar NetBS=
D?
(=C9ste es su =FAltimo aviso antes de que el programa empiece a modificar
sus discos.)
}
=20
+message Run_Postinst
+{Update files in /etc}
+
+message Postinst
+{Execute postinst (overwrite existing configuration)}
+
+message Etcupdate
+{Execute etcupdate (ask for every overwrite)}
+
+message Skip_postinst
+{Don't update any file on /etc}
+
message reinstallusure
{Se va a desempaquetar los conjuntos de distribuci=F3n de NetBSD
a un disco duro marcado como arrancable. Este procedimiento solamente
descarga y desempaqueta los conjuntos en un disco arrancable preparticiona=
do.
No pone nombre a los discos, ni actualiza los bloques de arranque, ni guar=
da
@@ -122,10 +134,13 @@
=BFRealmente quiere reinstalar los conjuntos de la distribuci=F3n NetBSD?
(=C9ste es su =FAltimo aviso antes de que el programa empiece a modificar
sus discos.)
}
=20
+message mount_failed
+{Mounting %s failed. Continue?
+}
=20
message nodisk
{No se ha podido encontrar ning=FAn disco duro para ser usado por NetBSD.
Se le volver=E1 a llevar al men=FA original.
}
@@ -699,11 +714,11 @@
{Creando nodos de dispositivo ...
}
=20
message badfs
{Parece que /dev/%s%c no es un sistema de archivos BSD o el fsck no ha sido
-correcto. La actualizaci=F3n ha sido interrumpida. (Error n=FAmero %d.)
+correcto (Error n=FAmero %d.) =BFContinuar?
}
=20
message rootmissing
{ el directorio ra=EDz objetivo no existe %s.
}
--MP_/joB4JmgFVA+c7H_120kwHN4
Content-Type: text/x-patch
Content-Transfer-Encoding: quoted-printable
Content-Disposition: attachment; filename=msg.mi.fr.diff
--- msg.mi.fr
+++ msg.mi.fr
@@ -107,10 +107,22 @@
Voulez-vous vraiment mettre =E0 jour NetBSD ?
(Ceci est le dernier avertissement avant que cette
proc=E9dure ne modifie votre disque.)
}
=20
+message Run_Postinst
+{Update files in /etc}
+
+message Postinst
+{Execute postinst (overwrite existing configuration)}
+
+message Etcupdate
+{Execute etcupdate (ask for every overwrite)}
+
+message Skip_postinst
+{Don't update any file on /etc}
+
message reinstallusure
{D=E9compressons maintenant les composants de NetBSD sur un disque dur
d=E9marrable.
Cette proc=E9dure va rapatrier et d=E9compresser les composants
sur un disque d=E9marrable d=E9j=E0 partitionnn=E9.
@@ -122,10 +134,14 @@
ou une mise =E0 jour avant de d=E9marrer cette proc=E9dure !
=20
Voulez-vous r=E9ellement r=E9installer les composants NetBSD ?
(Ceci est le dernier avertissement avant que cette proc=E9dure ne commence=
=E0
modifier vos disques.)
+}
+
+message mount_failed
+{Mounting %s failed. Continue?
}
=20
message nodisk
{Nous ne trouvons aucun disque utilisable par NetBSD. Vous allez retourner
au menu pr=E9c=E9dent.
@@ -740,13 +756,13 @@
}
=20
message badfs
{
/dev/%s%c ne semble pas =EAtre un syst=E8me de fichiers BSD ou bien
-la v=E9rification de son int=E9grit=E9 par fsck a =E9chou=E9.
+la v=E9rification de son int=E9grit=E9 par fsck a =E9chou=E9, code d'erreu=
r %d.
=20
-Interruption de la proc=E9dure de mise =E0 jour, code d'erreur %d.
+Souhaitez-vous continuer mise =E0 jour?
}
=20
message rootmissing
{
Impossible de trouver la racine du disque cible %s.
--MP_/joB4JmgFVA+c7H_120kwHN4
Content-Type: text/x-patch
Content-Transfer-Encoding: 7bit
Content-Disposition: attachment; filename=msg.mi.pl.diff
--- msg.mi.pl
+++ msg.mi.pl
@@ -106,10 +106,22 @@
przed rozpoczeciem! Czy napewno chcesz zaktualizowac NetBSD?
(Jest to ostatnie ostrzezenie zanim zacznie sie modyfikacja danych na
twoich dyskach.)
}
+message Run_Postinst
+{Update files in /etc}
+
+message Postinst
+{Execute postinst (overwrite existing configuration)}
+
+message Etcupdate
+{Execute etcupdate (ask for every overwrite)}
+
+message Skip_postinst
+{Don't update any file on /etc}
+
message reinstallusure
{Ok, rozpakujmy pakiety dystrybucyjne NetBSD na bootowalny twardy dysk.
Ta procedura tylko sciaga i rozpakowuje pakiety na pre-partycjonowany
bootowalny dysk. Nie nazywa dyskow, aktualizuje bootblokow, lub zapisuje
istniejacej konfiguracji. (Wyjdz i wybierz `instaluj' lub
@@ -119,10 +131,13 @@
Czy napewno chcesz przeinstalowac pakiety dystrybucjne NetBSD?
(Jest to ostatnie ostrzezenie zanim zacznie sie modyfikacja danych na
twoich dyskach.)
}
+message mount_failed
+{Mounting %s failed. Continue?
+}
message nodisk
{Nie moge znalezc zadnych twardych dyskow do uzycia z NetBSD. Zostaniesz
przeniesiony do menu glownego.
}
@@ -677,10 +692,14 @@
message badfs
{Wyglada na to, ze /dev/%s%c nie jest systemem plikow BSD albo nie powiodlo sie
jego sprawdzenie. Aktualizacja zostala przerwana. (Blad numer %d.)
}
+/* XXX: Translate:
+ * -not successful. The upgrade has been aborted. (Error number %d.)
+ * +not successful (Error number %d.). Try mounting it anyway?
+ */
message rootmissing
{ docelowy / jest zagubiony %s.
}
--MP_/joB4JmgFVA+c7H_120kwHN4
Content-Type: text/x-patch
Content-Transfer-Encoding: 7bit
Content-Disposition: attachment; filename=run.c.diff
--- run.c
+++ run.c
@@ -564,10 +564,86 @@
if (WIFSIGNALED(status)) {
*errstr = msg_string(MSG_Command_ended_on_signal);
return WTERMSIG(status);
}
return 0;
+}
+
+/*
+ * launch a program inside a subwindow, and report its return status when done
+ */
+int
+launch_command(int flags, const char *cmd)
+{
+ int status;
+ pid_t child, pid;
+ char **args;
+
+ args = make_argv(cmd);
+
+ /* Make curses save tty settings */
+ def_prog_mode();
+
+ if (logfp)
+ fflush(logfp);
+ if (script)
+ fflush(script);
+
+ child = fork();
+ if (child == -1) { /* error */
+ refresh();
+ return -1;
+ } else if (child == 0) { /* child */
+ /* silently stop curses */
+ endwin();
+
+ if (logfp) {
+ fprintf(logfp, "executing: %s\n", cmd);
+ fclose(logfp);
+ logfp = NULL;
+ }
+ if (script) {
+ fprintf(script, "%s\n", cmd);
+ fclose(script);
+ script = NULL;
+ }
+ if (strcmp(args[0], "cd") == 0 && strcmp(args[2], "&&") == 0) {
+ target_chdir_or_die(args[1]);
+ args += 3;
+ }
+ if (flags & RUN_XFER_DIR)
+ target_chdir_or_die(xfer_dir);
+ /*
+ * If target_prefix == "", the chroot will fail, but
+ * that's ok, since we don't need it then.
+ */
+ if (flags & RUN_CHROOT && *target_prefix()
+ && chroot(target_prefix()) != 0)
+ warn("chroot(%s) for %s", target_prefix(), *args);
+ else {
+ execvp(*args, args);
+ warn("execvp %s", *args);
+ }
+ _exit(EXIT_FAILURE);
+ // break; /* end of child */
+ } else { /* parent */
+ pid = waitpid(child, &status, 0);
+ }
+
+ if (logfp)
+ fflush(logfp);
+
+ reset_prog_mode();
+ free_argv(args);
+
+ if (WIFEXITED(status))
+ return WEXITSTATUS(status);
+ else if (WIFSIGNALED(status))
+ return WTERMSIG(status);
+
+ /* Never reached. */
+ return -1;
}
/*
* generic program runner.
* flags:
--MP_/joB4JmgFVA+c7H_120kwHN4
Content-Type: text/x-patch
Content-Transfer-Encoding: 7bit
Content-Disposition: attachment; filename=util.c.diff
--- util.c
+++ util.c
@@ -1061,10 +1061,12 @@
* otherwise /var/db/obsolete will only have current information
* from the base, comp, and etc sets.
*/
if (update && (set_status[SET_ETC] & SET_INSTALLED)) {
int oldsendmail;
+ int runpostinstall;
+
oldsendmail = run_program(RUN_DISPLAY | RUN_CHROOT |
RUN_ERROR_OK | RUN_PROGRESS,
"/usr/sbin/postinstall -s /.sysinst -d / check mailerconf");
if (oldsendmail == 1) {
msg_display(MSG_oldsendmail);
@@ -1072,12 +1074,38 @@
if (yesno) {
run_program(RUN_DISPLAY | RUN_CHROOT,
"/usr/sbin/postinstall -s /.sysinst -d / fix mailerconf");
}
}
- run_program(RUN_DISPLAY | RUN_CHROOT,
- "/usr/sbin/postinstall -s /.sysinst -d / fix");
+
+ runpostinstall = run_program(RUN_DISPLAY | RUN_CHROOT | RUN_ERROR_OK
+ | RUN_PROGRESS,
+ "/usr/sbin/postinstall -s /.sysinst -d / check");
+ if (runpostinstall == 1) {
+ process_menu(MENU_runpostinst, &runpostinstall);
+ /* runpostinstall:
+ * 0: postinstall fix
+ * 1: etcupdate
+ * 2: continue
+ * 3: abandon installation
+ */
+ switch(runpostinstall) {
+ case 0:
+ run_program(RUN_DISPLAY | RUN_CHROOT,
+ "/usr/sbin/postinstall -s /.sysinst -d / fix");
+ break;
+ case 1:
+ launch_command(RUN_CHROOT,
+ "/usr/sbin/etcupdate -s /.sysinst");
+ break;
+ case 2:
+ break;
+ case 3:
+ return -1;
+ break;
+ }
+ }
}
/* Configure the system */
if (set_status[SET_BASE] & SET_INSTALLED)
run_makedev();
--MP_/joB4JmgFVA+c7H_120kwHN4--
From: Martin Husemann <martin@duskware.de>
To: gnats-bugs@NetBSD.org
Cc:
Subject: Re: install/38571: sysinst runs "postinstall fix"
Date: Wed, 29 Feb 2012 10:10:35 +0100
On Wed, Feb 29, 2012 at 02:25:07AM +0000, Julian Djamil Fagir wrote:
> +message Postinst
> +{Execute postinst (overwrite existing configuration)}
This is not quite what it actually does, I'm not sure the user will be
always able to understand the implications and what the alternatives
are, so from a usability POV it is a bad question to ask.
Martin
From: Julian Djamil Fagir <gnrp@komkon2.de>
To: gnats-bugs@NetBSD.org
Cc:
Subject: Re: install/38571: sysinst runs "postinstall fix"
Date: Wed, 29 Feb 2012 11:49:04 +0100
--Sig_/xgjYv.xg/z53cusaRPacLLE
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: quoted-printable
Hi
> > +message Postinst
> > +{Execute postinst (overwrite existing configuration)}
> =20
> This is not quite what it actually does, I'm not sure the user will be
> always able to understand the implications and what the alternatives
> are, so from a usability POV it is a bad question to ask.
but what is a better one-line description of `postinstall fix`?
Btw, it should be named postinstall anyway (just now occurred to me).
Regards, Julian
--Sig_/xgjYv.xg/z53cusaRPacLLE
Content-Type: application/pgp-signature; name=signature.asc
Content-Disposition: attachment; filename=signature.asc
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (GNU/Linux)
iEYEARECAAYFAk9OAqAACgkQc7h7cu1Hpp5y9gCgsxtLFNv0UotpJv3lduHg6nrc
GhwAnRTLvxHRlz1ghrlFJVFHVg1u6B7l
=nh4q
-----END PGP SIGNATURE-----
--Sig_/xgjYv.xg/z53cusaRPacLLE--
From: Matthew Mondor <mm_lists@pulsar-zone.net>
To: gnats-bugs@NetBSD.org
Cc:
Subject: Re: install/38571: sysinst runs "postinstall fix"
Date: Wed, 29 Feb 2012 06:20:39 -0500
On Wed, 29 Feb 2012 10:50:04 +0000 (UTC)
Julian Djamil Fagir <gnrp@komkon2.de> wrote:
> but what is a better one-line description of `postinstall fix`?
> Btw, it should be named postinstall anyway (just now occurred to me).
Is that the actual execution of postinstall? If so, in my experience
it generally doesn't break /etc/, but it'll do things like delete
obsolete files, report if some users or groups must be added (as
postinstall cannot add those automatically yet), etc. It'll also
occasionally warn that some directories or files should be moved by the
administrator.
"Execute the post-installation/post-upgrade script"?
However, that also won't really fix /etc/ like adding new configuration
files or applying changes to them, for which etcupate is used.
In the current state of postinstall, the output has to also be
preserved so that the user can view it, perform any manual fixes
needed, and rerun the check/fix phases to verify that all is clean.
Of course, it'd probably be nice if the fix option could really
autonomously perform any fix including merging user and group
additions...
--
Matt
From: Alan Barrett <apb@cequrux.com>
To: gnats-bugs@NetBSD.org
Cc:
Subject: Re: install/38571: sysinst runs "postinstall fix"
Date: Wed, 29 Feb 2012 13:35:03 +0200
On Wed, 29 Feb 2012, Julian Djamil Fagir wrote:
> the attached patch fixes that issue for me.
Thsnk you for working on this.
> On the one hand, it introduces a
> new function that exits curses, executes a command and afterwards returns.
> This function is used to execute etcupdate(8) in case the user wants it.
That's fine in principle. I haven't checked the implementation.
> + runpostinstall = run_program(RUN_DISPLAY | RUN_CHROOT | RUN_ERROR_OK
> + | RUN_PROGRESS,
> + "/usr/sbin/postinstall -s /.sysinst -d / check");
> + if (runpostinstall == 1) {
If run_program returns the program's exit status, then any non-zero
value means that something needs to be fixed.
> + process_menu(MENU_runpostinst, &runpostinstall);
> + /* runpostinstall:
> + * 0: postinstall fix
> + * 1: etcupdate
> + * 2: continue
> + * 3: abandon installation
> + */
Will the output from postinstall still be visible when the user
has to choose a menu option?
Also, I think we could have "re-run postinstall check" and "run
a shell to fix problems manually" as additional options in thie
menu, and loop here until the user is happy. (Maybe they want to
run postinstall check, fix something in a shell, postinstall check
again, etcupdate, postinstall check again, fix something else in a
shell, ...)
I usually want to run "postinstall fix SOME_KEYWORDS" and not
"postinstall fix OTHER_KEYWORDS". With enough work, we could
dynamically build a menu using a list of keywords parsed from the
output of "postinstall check", and descriptions parsed from the
output of "postinstall list".
--apb (Alan Barrett)
From: Alan Barrett <apb@cequrux.com>
To: gnats-bugs@NetBSD.org
Cc:
Subject: Re: install/38571: sysinst runs "postinstall fix"
Date: Wed, 29 Feb 2012 13:41:01 +0200
On Wed, 29 Feb 2012, Martin Husemann wrote:
> On Wed, Feb 29, 2012 at 02:25:07AM +0000, Julian Djamil Fagir wrote:
> > +message Postinst
> > +{Execute postinst (overwrite existing configuration)}
>
> This is not quite what it actually does,
Yes, the message could be improved. Perhaps "Run 'postinstall fix' to
fix the problems reported above".
> I'm not sure the user will be always able to understand the
> implications and what the alternatives are, so from a usability
> POV it is a bad question to ask.
"postinstall fix" is dangerous, in my opinion, because it
overwrites changes that the user may have made to /etc/rc.d/*,
/etc/daily and friends, and probably some other configuration
files. Asking the question may be bad for usability, but simply
overwriting the user's changes is also bad.
--apb (Alan Barrett)
From: Julian Djamil Fagir <gnrp@komkon2.de>
To: gnats-bugs@NetBSD.org
Cc:
Subject: Re: install/38571: sysinst runs "postinstall fix"
Date: Wed, 29 Feb 2012 12:43:38 +0100
Hi,
> > but what is a better one-line description of `postinstall fix`?
> > Btw, it should be named postinstall anyway (just now occurred to me).
>
> Is that the actual execution of postinstall? If so, in my experience
> it generally doesn't break /etc/, but it'll do things like delete
> obsolete files, report if some users or groups must be added (as
> postinstall cannot add those automatically yet), etc. It'll also
> occasionally warn that some directories or files should be moved by the
> administrator.
>
> "Execute the post-installation/post-upgrade script"?
but that's not a clear distinction from etcupdate(8) imho.
> However, that also won't really fix /etc/ like adding new configuration
> files or applying changes to them, for which etcupate is used.
>
> In the current state of postinstall, the output has to also be
> preserved so that the user can view it, perform any manual fixes
> needed, and rerun the check/fix phases to verify that all is clean.
> Of course, it'd probably be nice if the fix option could really
> autonomously perform any fix including merging user and group
> additions...
Currently, there's a postinstall(8) run when `postinstall check` fails after
an upgrade. According to its manpage, it replaces existing files with the new
one and does not attempt to merge them.
Doesn't the documentation match the actual program?
Regards, Julian
From: Matthew Mondor <mm_lists@pulsar-zone.net>
To: gnats-bugs@NetBSD.org
Cc:
Subject: Re: install/38571: sysinst runs "postinstall fix"
Date: Wed, 29 Feb 2012 07:12:07 -0500
On Wed, 29 Feb 2012 11:45:04 +0000 (UTC)
Julian Djamil Fagir <gnrp@komkon2.de> wrote:
> Currently, there's a postinstall(8) run when `postinstall check` fails after
> an upgrade. According to its manpage, it replaces existing files with the new
> one and does not attempt to merge them.
> Doesn't the documentation match the actual program?
Probably at least for files in rc.d like Alan noted, I guess. I only
made additions to rc.d rather editing existing files, so have never
noticed a problem using it after source upgrades.
--
Matt
From: Julian Djamil Fagir <gnrp@komkon2.de>
To: gnats-bugs@NetBSD.org
Cc:
Subject: Re: install/38571: sysinst runs "postinstall fix"
Date: Wed, 29 Feb 2012 14:05:59 +0100
--MP_/5XsSkzMyzyEn1mIEIo_l7K4
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Content-Disposition: inline
Hi,
> > On the one hand, it introduces a
> > new function that exits curses, executes a command and afterwards
> > returns. This function is used to execute etcupdate(8) in case the user
> > wants it.
>
> That's fine in principle. I haven't checked the implementation.
the implementation is what I'm most concerned about. ;-)
> > + runpostinstall = run_program(RUN_DISPLAY | RUN_CHROOT
> > | RUN_ERROR_OK
> > + | RUN_PROGRESS,
> > + "/usr/sbin/postinstall -s /.sysinst
> > -d / check");
> > + if (runpostinstall == 1) {
>
> If run_program returns the program's exit status, then any non-zero
> value means that something needs to be fixed.
Fixed.
> Will the output from postinstall still be visible when the user
> has to choose a menu option?
No, it's not. It's difficult to do that, anyway, as it isn't as ordered as it
should be for piping it somewhere.
menuc doesn't really have methods to display the output of a program and a
window afterwards.
> Also, I think we could have "re-run postinstall check" and "run
> a shell to fix problems manually" as additional options in thie
> menu, and loop here until the user is happy. (Maybe they want to
> run postinstall check, fix something in a shell, postinstall check
> again, etcupdate, postinstall check again, fix something else in a
> shell, ...)
What about the attached patch?
It loops though the process, until `postinstall check` is ok or the user
explicitly stops.
> I usually want to run "postinstall fix SOME_KEYWORDS" and not
> "postinstall fix OTHER_KEYWORDS". With enough work, we could
> dynamically build a menu using a list of keywords parsed from the
> output of "postinstall check", and descriptions parsed from the
> output of "postinstall list".
This is difficult and a heck of work. As sysinst should be replaced anyway, I
wonder if it's really worth the time implementing this.
Regards, Julian
--MP_/5XsSkzMyzyEn1mIEIo_l7K4
Content-Type: text/x-patch
Content-Transfer-Encoding: 7bit
Content-Disposition: attachment; filename=util.c.diff
--- util.c
+++ util.c
@@ -571,11 +571,11 @@
if (silent)
msg_display("");
else {
umount_mnt2();
msg_display(MSG_cd_path_not_found);
- msg_display_add("\r\n\r\n");
+ process_menu(MENU_ok, NULL);
}
/* ask for paths on the CD */
process_menu(MENU_cdromsource, NULL);
@@ -1061,10 +1061,12 @@
* otherwise /var/db/obsolete will only have current information
* from the base, comp, and etc sets.
*/
if (update && (set_status[SET_ETC] & SET_INSTALLED)) {
int oldsendmail;
+ int runpostinstall;
+
oldsendmail = run_program(RUN_DISPLAY | RUN_CHROOT |
RUN_ERROR_OK | RUN_PROGRESS,
"/usr/sbin/postinstall -s /.sysinst -d / check mailerconf");
if (oldsendmail == 1) {
msg_display(MSG_oldsendmail);
@@ -1072,12 +1074,49 @@
if (yesno) {
run_program(RUN_DISPLAY | RUN_CHROOT,
"/usr/sbin/postinstall -s /.sysinst -d / fix mailerconf");
}
}
- run_program(RUN_DISPLAY | RUN_CHROOT,
- "/usr/sbin/postinstall -s /.sysinst -d / fix");
+
+ runpostinstall = run_program(RUN_DISPLAY | RUN_CHROOT | RUN_ERROR_OK
+ | RUN_PROGRESS,
+ "/usr/sbin/postinstall -s /.sysinst -d / check");
+ while (runpostinstall != 0) {
+ process_menu(MENU_runpostinst, &runpostinstall);
+ /* runpostinstall:
+ * 0: postinstall fix
+ * 1: etcupdate
+ * 2: Run shell
+ * 3: continue
+ * 4: abandon installation
+ */
+ switch(runpostinstall) {
+ case 0:
+ run_program(RUN_DISPLAY | RUN_CHROOT,
+ "/usr/sbin/postinstall -s /.sysinst -d / fix");
+ break;
+ case 1:
+ launch_command(RUN_CHROOT,
+ "/usr/sbin/etcupdate -s /.sysinst");
+ break;
+ case 2:
+ launch_command(RUN_CHROOT,
+ "/bin/sh");
+ break;
+ case 3:
+ runpostinstall = 0;
+ break;
+ case 4:
+ return -1;
+ break;
+ }
+
+ if (runpostinstall != 0)
+ run_program(RUN_DISPLAY | RUN_CHROOT | RUN_ERROR_OK
+ | RUN_PROGRESS,
+ "/usr/sbin/postinstall -s /.sysinst -d / check");
+ }
}
/* Configure the system */
if (set_status[SET_BASE] & SET_INSTALLED)
run_makedev();
--MP_/5XsSkzMyzyEn1mIEIo_l7K4--
From: David Laight <david@l8s.co.uk>
To: gnats-bugs@NetBSD.org
Cc:
Subject: Re: install/38571: sysinst runs "postinstall fix"
Date: Wed, 29 Feb 2012 18:22:52 +0000
On Wed, Feb 29, 2012 at 02:25:07AM +0000, Julian Djamil Fagir wrote:
> The following reply was made to PR install/38571; it has been noted by GNATS.
>
> From: Julian Djamil Fagir <gnrp@komkon2.de>
> To: gnats-bugs@gnats.netbsd.org
> Cc:
> Subject: Re: install/38571: sysinst runs "postinstall fix"
> Date: Wed, 29 Feb 2012 02:30:29 +0100
>
> --MP_/joB4JmgFVA+c7H_120kwHN4
> Content-Type: text/plain; charset=US-ASCII
> Content-Transfer-Encoding: 7bit
> Content-Disposition: inline
>
> Hi,
>
> the attached patch fixes that issue for me. On the one hand, it introduces a
> new function that exits curses, executes a command and afterwards returns.
> This function is used to execute etcupdate(8) in case the user wants it.
Why is that needed, IIRC there is already stuff to run a program and
display its output in the curses window.
(Or am I just thinking of the code that displays output to /dev/console.)
David
--
David Laight: david@l8s.co.uk
From: Alan Barrett <apb@netbsd.org>
To: gnats-bugs@netbsd.org
Cc:
Subject: Re: install/38571: sysinst runs "postinstall fix"
Date: Wed, 29 Feb 2012 20:42:16 +0200
On Wed, 29 Feb 2012, David Laight wrote:
>> the attached patch fixes that issue for me. On the one hand, it
>> introduces a new function that exits curses, executes a command
>> and afterwards returns. This function is used to execute
>> etcupdate(8) in case the user wants it.
>
> Why is that needed, IIRC there is already stuff to run a program
> and display its output in the curses window.
Yes, we have that, but does it work for interactive programs?
AFAIK we use it only for programs that spew output without reading
anything from the keyboard.
--apb (Alan Barrett)
From: Julian Fagir <gnrp@komkon2.de>
To: gnats-bugs@NetBSD.org
Cc:
Subject: Re: install/38571: sysinst runs "postinstall fix"
Date: Wed, 29 Feb 2012 20:25:31 +0100
Hi,
> >> the attached patch fixes that issue for me. On the one hand, it
> >> introduces a new function that exits curses, executes a command
> >> and afterwards returns. This function is used to execute
> >> etcupdate(8) in case the user wants it.
> >
> > Why is that needed, IIRC there is already stuff to run a program
> > and display its output in the curses window.
>
> Yes, we have that, but does it work for interactive programs?
> AFAIK we use it only for programs that spew output without reading
> anything from the keyboard.
yep, and I derived the code for this function from the existing one.
I don't remember exactly anymore, but there were cases where the top three
lines occupying the view of the command were breaking things.
I think it was when looking at files or diffs, then you never could see the
top three lines when scrolling a page, but wouldn't bet on it.
Regards, Julian
From: matthew green <mrg@eterna.com.au>
To: gnats-bugs@NetBSD.org
Cc: install-manager@netbsd.org, gnats-admin@netbsd.org,
netbsd-bugs@netbsd.org, apb@cequrux.com
Subject: re: install/38571: sysinst runs "postinstall fix"
Date: Thu, 01 Mar 2012 10:54:56 +1100
i'm curious what local configuration postinstall breaks. apb, you had
the original report can you expand on this?
to me, that sounds like a bug itself that needs to be fixed.
wrt the message, "update configuration" sounds better than "overwrite"
to me. (overwrite is a bug, to me.)
.mrg.
From: Julian Djamil Fagir <gnrp@komkon2.de>
To: gnats-bugs@NetBSD.org
Cc:
Subject: Re: install/38571: sysinst runs "postinstall fix"
Date: Sun, 16 Sep 2012 02:02:47 +0200
--Sig_/_jnx9sQ+mjDX8nI_MQzq/pt
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: quoted-printable
Hi,
> i'm curious what local configuration postinstall breaks. apb, you had
> the original report can you expand on this?
> =20
> to me, that sounds like a bug itself that needs to be fixed.
a quick check (only looking at code, no real test) revealed that the
following files at least will be replaced with those from the sets:
* /etc/defaults/{rc.conf,daily.conf,monthly.conf,weekly.conf,
security.conf,pf.boot.conf}
* /dev/{MAKEDEV,MAKEDEV.local}
* /etc/mtree/{NetBSD.dist,special}
* /etc/namedb/root.cache
* /etc/{daily,weekly,monthly,security}
* /etc/pf.os
* /etc/{rc,rc.subr,rc.shutdown}
* /etc/rc.d/{DAEMON,DISKS,LOGIN,NETWORKING,SERVERS,accounting,altqd,
amd,apmd,bluetooth,bootconf.sh,bootparams,ccd,cgd,cleartmp,cron,
dhclient,dhcpcd,dhcpd,dhcrelay,dmesg,downinterfaces,envsys,fsck,
fsck_root,ftp_proxy,ftpd,gpio,hostapd,httpd,identd,ifwatchd,inetd,
ipfilter,ipfs,ipmon,ipnat,ipsec,irdaattach,iscsi_target,isdnd,
isibootd,kdc,ldconfig,ldpd,local,lpd,lvm,makemandb,mdnsd,mixerctl,
mopd,motd,mountall,mountcritlocal,mountcritremote,mountd,moused,
mrouted,named,ndbootd,network,newsyslog,nfsd,nfslocking,npf,ntpd,
ntpdate,perusertmp,pf,pf_boot,pflogd,postfix,powerd,ppp,pwcheck,
quota,racoon,rpcbind,raidframe,raidframeparity,random_seed,rarpd,
rbootd,rndctl,root,route6d,routed,rtadvd,rtclocaltime,rtsold,rwho,
savecore,screenblank,securelevel,sshd,staticroute,swap1,swap2,
sysctl,sysdb,syslogd,timed,tpctl,ttys,veriexec,virecover,wdogctl,
wpa_supplicant,wscons,wsmoused,ypbind,yppasswdd,ypserv}
* /etc/mailer.conf
Regards, Julian
--Sig_/_jnx9sQ+mjDX8nI_MQzq/pt
Content-Type: application/pgp-signature; name=signature.asc
Content-Disposition: attachment; filename=signature.asc
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (GNU/Linux)
iEYEARECAAYFAlBVFygACgkQc7h7cu1Hpp6CigCbB165UlLksOHEDSKH6HQPjngJ
XesAnRvnpgjXc9GzddWo/uf5pxAiJhFw
=9u7l
-----END PGP SIGNATURE-----
--Sig_/_jnx9sQ+mjDX8nI_MQzq/pt--
From: David Holland <dholland-bugs@netbsd.org>
To: gnats-bugs@NetBSD.org
Cc:
Subject: Re: install/38571: sysinst runs "postinstall fix"
Date: Sun, 16 Sep 2012 01:02:04 +0000
On Sun, Sep 16, 2012 at 12:05:04AM +0000, Julian Djamil Fagir wrote:
> * /etc/defaults/{rc.conf,daily.conf,monthly.conf,weekly.conf,
> security.conf,pf.boot.conf}
> * /dev/{MAKEDEV,MAKEDEV.local}
> * /etc/mtree/{NetBSD.dist,special}
> * /etc/namedb/root.cache
> * /etc/{daily,weekly,monthly,security}
> * /etc/pf.os
> * /etc/{rc,rc.subr,rc.shutdown}
> * /etc/rc.d/*
> * /etc/mailer.conf
Right, the only files in here that are supposed to be edited locally
are MAKEDEV.local and mailer.conf; postinstall shouldn't be trashing
those. (I believe all it will do to mailer.conf is migrate from base
sendmail to base postfix. No idea about MAKEDEV.local.)
The rest of that stuff is no more editable than, say, /sbin/mount is,
and these files are only in /etc for historical reasons. Arguably they
should just be updated automatically during installworld and not left
to postinstall at all.
--
David A. Holland
dholland@netbsd.org
(Contact us)
$NetBSD: query-full-pr,v 1.39 2013/11/01 18:47:49 spz Exp $
$NetBSD: gnats_config.sh,v 1.8 2006/05/07 09:23:38 tsutsui Exp $
Copyright © 1994-2007
The NetBSD Foundation, Inc. ALL RIGHTS RESERVED.
Home