From rcrit at greyoak.com  Thu Apr  8 14:55:22 2004
From: rcrit at greyoak.com (Rob Crittenden)
Date: Thu, 8 Apr 2004 14:55:22 -0400 (EDT)
Subject: script patch
Message-ID: <Pine.LNX.4.44.0404081448240.23861-200000@zippy.greyoak.com>

Hi.

I was working on a multi-cd installer that uses a script to copy some of
the files from the cds. The environment variable SETUP_CDROMPATH wasn't
always being set because the installer always outputs the top of the cdrom
list. So when you're on cd 2 it tries to print the mount point for cd 1
which is NULL when you're working on cd2.

I added a cdromid attribute to the script command so one can say which
cdrom the script is operating against. So if a cdromid attrbute exists
I save a pointer to the head of the list, then seek to cdromid in the
cdrom list, then call the script and finally restore the pointer when the
script exits.

While I was at it I added 2 more attributes: size and message.

Size lets you indicate the amount of data that the script installs itself.

Message lets you replace the "Running script" message with something slightly
more useful to the end-user.

The installer I was working on used Microsoft cabinet files. Rather than
porting cabextract to a setup plugin I call it from a shell script instead.
Since the majority (90%) of the files are in .cab files the total installed
progress bar was about 5% during 95% of the installation. So instead with
the size message at the end of each installation script the total progress
goes up a bit.

A sample entry looks like:

  <script cdromid="disc1" size="3504531" message="Installing Game.cab">
   sh setup.data/install-cabGame.sh
  </script>

I wasn't sure where to submit this so I figured I'd start here rather than
bugging Ryan about it, he seems busy enough.

rob
-------------- next part --------------
Index: copy.c
===================================================================
RCS file: /cvs/cvsroot/loki_setup/copy.c,v
retrieving revision 1.74
diff -u -r1.74 copy.c
--- copy.c	1 Mar 2004 06:29:42 -0000	1.74
+++ copy.c	8 Apr 2004 18:39:13 -0000
@@ -729,16 +729,36 @@
 }
 
 int copy_script(install_info *info, xmlNodePtr node, const char *script, const char *dest,
-				int (*update)(install_info *info, const char *path, size_t progress, size_t size, const char *current))
+				int (*update)(install_info *info, const char *path, size_t progress, size_t size, const char *current), const char *from_cdrom, const char *msg)
 {
+    struct cdrom_elem *cdrom;
+    struct cdrom_elem *cdrom_start;
+    int rc;
+
 	if ( corrupts ) { /* Don't run any scripts while restoring files */
 		return 0;
 	}
 	if ( update ) {
-		if ( ! update(info, _("Running script"), 0, 0, current_option_txt) )
+		if (msg != NULL)
+			rc = update(info, msg, 0, 0, current_option_txt);
+		else
+			rc = update(info, _("Running script"), 0, 0, current_option_txt);
+		if ( !rc ) 
 			return 0;
 	}
-    return(run_script(info, script, -1, 1));
+
+    cdrom_start = info->cdroms_list;
+    while ( from_cdrom != NULL && info->cdroms_list ) {
+        cdrom = info->cdroms_list;
+        if (!strcmp(cdrom->id, from_cdrom)) {
+            break;
+        }
+        info->cdroms_list = cdrom->next;
+    }
+
+    rc = run_script(info, script, -1, 1);
+    info->cdroms_list = cdrom_start;
+    return rc;
 }
 
 ssize_t copy_node(install_info *info, xmlNodePtr node, const char *dest,
@@ -819,9 +839,16 @@
 					size += copied;
 				}
 			} else if ( strcmp(node->name, "script") == 0 ) {
-				copy_script(info, node,
+				const char *sz= xmlGetProp(node, "size");
+				const char *msg= xmlGetProp(node, "message");
+				int rc;
+				rc = copy_script(info, node,
 					    xmlNodeListGetString(info->config, node->childs, 1),
-					    path, update);
+					    path, update, from_cdrom, msg);
+				if (rc == 0 && sz) {
+					info->installed_bytes += atoi(sz);
+					size += atoi(sz);
+				}
 			}
 		}
         /* Do not handle exclusive elements here; it gets called multiple times else */
Index: README.xml
===================================================================
RCS file: /cvs/cvsroot/loki_setup/README.xml,v
retrieving revision 1.57
diff -u -r1.57 README.xml
--- README.xml	18 Feb 2004 03:56:55 -0000	1.57
+++ README.xml	8 Apr 2004 18:39:16 -0000
@@ -760,6 +760,18 @@
             string of this attribute. See 'About localization' below for
             more details.
 
+ size       If the script handles the installation of some files and you want
+            the progress bar to reflect it, indicate the size here. The size
+            is updated when the script finishes. The unit is in bytes.
+
+ message    The message to display to the user. By default this is
+            "Running script"
+
+ cdromid    When doing a multi-cd installation and running scripts then
+            SETUP_CDROMPATH may not be set as expected. To ensure that it is
+            set properly indicate which CD the script is executing against
+            if it is copying files from a CD.
+
 It is also possible to filter the scripts on system characteristics, through
 the use of the "arch", "glibc" and "distro" attributes, that work just like
 for the "files" element.

From megastep at megastep.org  Thu Apr  8 19:00:56 2004
From: megastep at megastep.org (Stephane Peter)
Date: Thu, 8 Apr 2004 16:00:56 -0700
Subject: [lokisetup] script patch
In-Reply-To: <Pine.LNX.4.44.0404081448240.23861-200000@zippy.greyoak.com>
References: <Pine.LNX.4.44.0404081448240.23861-200000@zippy.greyoak.com>
Message-ID: <9813AAF6-89B0-11D8-B5B6-000A959DD562@megastep.org>

Hi Rob,

Thanks for your patch - I'm looking at it and will probably merge parts 
if not all of it in the main CVS tree. You may want to contact me 
directly in the future for patch submissions, too. I'm maintaining 
loki_setup at the moment, though Ryan is an occasional contributor, as 
you know he is very busy ;-)

On Apr 8, 2004, at 11:55 AM, Rob Crittenden wrote:

> Hi.
>
> I was working on a multi-cd installer that uses a script to copy some 
> of
> the files from the cds. The environment variable SETUP_CDROMPATH wasn't
> always being set because the installer always outputs the top of the 
> cdrom
> list. So when you're on cd 2 it tries to print the mount point for cd 1
> which is NULL when you're working on cd2.
>
> I added a cdromid attribute to the script command so one can say which
> cdrom the script is operating against. So if a cdromid attrbute exists
> I save a pointer to the head of the list, then seek to cdromid in the
> cdrom list, then call the script and finally restore the pointer when 
> the
> script exits.
>
> While I was at it I added 2 more attributes: size and message.
>
> Size lets you indicate the amount of data that the script installs 
> itself.
>
> Message lets you replace the "Running script" message with something 
> slightly
> more useful to the end-user.
>
> The installer I was working on used Microsoft cabinet files. Rather 
> than
> porting cabextract to a setup plugin I call it from a shell script 
> instead.
> Since the majority (90%) of the files are in .cab files the total 
> installed
> progress bar was about 5% during 95% of the installation. So instead 
> with
> the size message at the end of each installation script the total 
> progress
> goes up a bit.
>
> A sample entry looks like:
>
>   <script cdromid="disc1" size="3504531" message="Installing Game.cab">
>    sh setup.data/install-cabGame.sh
>   </script>
>
> I wasn't sure where to submit this so I figured I'd start here rather 
> than
> bugging Ryan about it, he seems busy enough.
>
> rob
> <copy.diff>
--
St?phane Peter
megastep at megastep.org



From archlyn at comcast.net  Mon Apr 12 20:45:07 2004
From: archlyn at comcast.net (Robert Townsand)
Date: Mon, 12 Apr 2004 20:45:07 -0400
Subject: UT2003 troubles
Message-ID: <407B3813.30709@comcast.net>

I'm not entirely sure if this is the proper place for this, but I need 
help.  I've intalled UT2003 in windows. Works like a dream.   If I use 
those same disks in Linux, I get the following error, on a random file:

failed to uncompress file.
./setup.sh: line 214: 19848 Aborted                 "$setup" "$@"
./setup.sh: line 214: 24167 Segmentation fault      "$setup" "$@"
The setup program seems to have failed on x86/glibc-2.1

I've asked this on several linux BBs, googled for it, but still have no 
answer.

System specs:
AMD Athlon XP 2400+
512MBs RAM
GF4 TI 4200
SBLive 5.1
Slackware-current (updated regularly by swaret) w/ kernel 2.6.5

Any ideas?


From relic2k at canada.com  Tue Apr 13 05:37:33 2004
From: relic2k at canada.com (relic2k at canada.com)
Date: Tue, 13 Apr 2004 02:37:33 -0700 (PDT)
Subject: [lokisetup] UT2003 troubles
Message-ID: <20040413023734.26716.h021.c009.wm@mail.canada.com.criticalpath.net>

On Mon, 12 Apr 2004 20:45:07 -0400, Robert Townsand
wrote:

> 
> I'm not entirely sure if this is the proper place for
> this, but I need 
> help.  I've intalled UT2003 in windows. Works like a
> dream.   If I use 
> those same disks in Linux, I get the following error,
> on a random file:
> 
> failed to uncompress file.
> ./setup.sh: line 214: 19848 Aborted                
> "$setup" "$@"
> ./setup.sh: line 214: 24167 Segmentation fault     
> "$setup" "$@"
> The setup program seems to have failed on
x86/glibc-2.1
> 
> I've asked this on several linux BBs, googled for it,
> but still have no 
> answer.
> 
> System specs:
> AMD Athlon XP 2400+
> 512MBs RAM
> GF4 TI 4200
> SBLive 5.1
> Slackware-current (updated regularly by swaret) w/
> kernel 2.6.5
> 
> Any ideas?

I too started having this problem as of yesterday. I am
also not able to update via Loki Update. It errors out
complaining about the newer Glibc 2.3.3 TLS verion that
comes with MDL 10.0. I am wondering if perhaps my
UT2003 CDs are getting corrupted from an installations
that I have been doing to get the game updated properly
and trying to figure out how to get UT2003 updated to
the proper version with out it giving me an error all
the time. I need to get my game patched so that I can
work on my own MOD installers that I have created for
UT2003. This is just a thought but perhaps your CD's
are going corrupt too. Perhaps you should make back up
copies before anything gets worse. 


From relic2k at canada.com  Tue Apr 13 05:37:33 2004
From: relic2k at canada.com (relic2k at canada.com)
Date: Tue, 13 Apr 2004 02:37:33 -0700 (PDT)
Subject: [lokisetup] UT2003 troubles
Message-ID: <20040413023734.26716.h021.c009.wm@mail.canada.com.criticalpath.net>

On Mon, 12 Apr 2004 20:45:07 -0400, Robert Townsand
wrote:

> 
> I'm not entirely sure if this is the proper place for
> this, but I need 
> help.  I've intalled UT2003 in windows. Works like a
> dream.   If I use 
> those same disks in Linux, I get the following error,
> on a random file:
> 
> failed to uncompress file.
> ./setup.sh: line 214: 19848 Aborted                
> "$setup" "$@"
> ./setup.sh: line 214: 24167 Segmentation fault     
> "$setup" "$@"
> The setup program seems to have failed on
x86/glibc-2.1
> 
> I've asked this on several linux BBs, googled for it,
> but still have no 
> answer.
> 
> System specs:
> AMD Athlon XP 2400+
> 512MBs RAM
> GF4 TI 4200
> SBLive 5.1
> Slackware-current (updated regularly by swaret) w/
> kernel 2.6.5
> 
> Any ideas?

I too started having this problem as of yesterday. I am
also not able to update via Loki Update. It errors out
complaining about the newer Glibc 2.3.3 TLS verion that
comes with MDL 10.0. I am wondering if perhaps my
UT2003 CDs are getting corrupted from an installations
that I have been doing to get the game updated properly
and trying to figure out how to get UT2003 updated to
the proper version with out it giving me an error all
the time. I need to get my game patched so that I can
work on my own MOD installers that I have created for
UT2003. This is just a thought but perhaps your CD's
are going corrupt too. Perhaps you should make back up
copies before anything gets worse. 


From archlyn at comcast.net  Tue Apr 13 18:10:26 2004
From: archlyn at comcast.net (Robert Townsand)
Date: Tue, 13 Apr 2004 18:10:26 -0400
Subject: Deus Ex 1?
Message-ID: <407C6552.6050507@comcast.net>

Hello again. I finally got UT2003 working again. Turns out it was a bad 
disc  >:( I re-burned it and the game installed just fine :)

Now onto my question:
I was poking around Loki's site and noticed that Deus Ex 1 was "in 
development"

Was it ever released to the public? If so, can it be legally downloaded?


From chunky at icculus.org  Tue Apr 13 18:33:01 2004
From: chunky at icculus.org (Chunky Kibbles)
Date: Tue, 13 Apr 2004 18:33:01 -0400
Subject: [lokisetup] Deus Ex 1?
In-Reply-To: <407C6552.6050507@comcast.net>
References: <407C6552.6050507@comcast.net>
Message-ID: <20040413223301.GC16880@gamehenge.icculus.org>

On Tue, Apr 13, 2004 at 06:10:26PM -0400, Robert Townsand wrote:
> Hello again. I finally got UT2003 working again. Turns out it was a bad 
> disc  >:( I re-burned it and the game installed just fine :)

Last I checked, ut2k3 wasn't an exercise in CDs you can legally burn.

> Now onto my question:
> I was poking around Loki's site and noticed that Deus Ex 1 was "in 
> development"
> 
> Was it ever released to the public? If so, can it be legally downloaded?

No, and no. Doubly so since you're a warez monkey.

Sorry,
Gary (-;


From zakk at timedoctor.org  Tue Apr 13 18:34:30 2004
From: zakk at timedoctor.org (Zachary J. Slater)
Date: Tue, 13 Apr 2004 15:34:30 -0700
Subject: [lokisetup] Deus Ex 1?
In-Reply-To: <407C6552.6050507@comcast.net>
References: <407C6552.6050507@comcast.net>
Message-ID: <20040413223429.GA28566@timedoctor.org>

With regard to Deus Ex 1: No and No.
-- 
-Zachary J. Slater
zakk at timedoctor.org


From ludwig.nussel at gmx.de  Sat Apr 17 12:53:00 2004
From: ludwig.nussel at gmx.de (Ludwig Nussel)
Date: Sat, 17 Apr 2004 18:53:00 +0200
Subject: [PATCH] uninstall.c broken
Message-ID: <20040417165300.GA13929@defiant.wachendorf.lan>

Hi,

loki_uninstall doesn't compile. Fix:

Index: loki_setup/uninstall.c
===================================================================
--- loki_setup.orig/uninstall.c	2004-04-17 18:09:14.000000000 +0200
+++ loki_setup/uninstall.c	2004-04-17 18:33:16.149849255 +0200
@@ -49,7 +49,7 @@
 		  "  -v | --version   : Get version information.\n"
 		  "  product [component] : Uninstalls the specified product, or its subcomponent.\n"
 		  "  product < -l | --list > : List installed subcomponents of product.\n"
-		  ), argv[0]);
+		  ), argv0);
 }
 
 static void log_file(const char *name, const char *reason)

cu
Ludwig

-- 
(o_  Ludwig.Nussel at gmx.de
//\  PGP Key ID: FF8135CE
V_/_ ICQ:        52166811


From ludwig.nussel at gmx.de  Sat Apr 17 15:34:42 2004
From: ludwig.nussel at gmx.de (Ludwig Nussel)
Date: Sat, 17 Apr 2004 21:34:42 +0200
Subject: [PATCH] subfs
Message-ID: <20040417193442.GA1862@defiant.wachendorf.lan>

Hi,

SUSE 9.1 uses subfs to automatically mount removable media.
Unfortunately subfs mounted CD-ROMs don't appear in /etc/mtab so
loki-setup doesn't find them. A subfs line in /etc/fstab looks like
this:

/dev/cdrom /media/cdrom subfs fs=cdfss,ro,procuid,nosuid,nodev,exec,iocharset=utf8 0 0

ut2004 installation works with the following patch:

Index: loki_setup/detect.c
===================================================================
--- loki_setup.orig/detect.c	2004-04-17 18:27:01.456734874 +0200
+++ loki_setup/detect.c	2004-04-17 18:27:12.427395593 +0200
@@ -580,17 +580,20 @@
     mountfp = setmntent(MOUNTS_FILE, "r" );
     if( mountfp != NULL ) {
         while( (mntent = getmntent( mountfp )) != NULL && num_cdroms < SETUP_MAX_DRIVES){
-            char *tmp, mntdev[1024], mnt_type[32];
+            char *tmp, mntdev[1024], mnt_type[1024];
 
             if ( strncmp(mntent->mnt_fsname, DEVICE_FLOPPY, strlen(DEVICE_FLOPPY)) == 0)
                 continue;
 
-            strcpy(mntdev, mntent->mnt_fsname);
-            strcpy(mnt_type, mntent->mnt_type);
+#define XXXstrcpy(d,s) \
+	    do { strncpy(d,s,sizeof(d)); d[sizeof(d)-1] = '\0'; } while(0);
+
+            XXXstrcpy(mntdev, mntent->mnt_fsname);
+            XXXstrcpy(mnt_type, mntent->mnt_type);
             if ( strcmp(mnt_type, MNTTYPE_SUPER) == 0 ) {
                 tmp = strstr(mntent->mnt_opts, "fs=");
                 if ( tmp ) {
-                    strcpy(mnt_type, tmp+strlen("fs="));
+                    XXXstrcpy(mnt_type, tmp+strlen("fs="));
                     tmp = strchr(mnt_type, ',');
                     if ( tmp ) {
                         *tmp = '\0';
@@ -598,7 +601,7 @@
                 }
                 tmp = strstr(mntent->mnt_opts, "dev=");
                 if ( tmp ) {
-                    strcpy(mntdev, tmp+strlen("dev="));
+                    XXXstrcpy(mntdev, tmp+strlen("dev="));
                     tmp = strchr(mntdev, ',');
                     if ( tmp ) {
                         *tmp = '\0';
@@ -606,14 +609,28 @@
                 }
             }
 
+	    if ( strcmp(mnt_type, "subfs") == 0 ) {
+                tmp = strstr(mntent->mnt_opts, "fs=");
+                if ( tmp ) {
+                    XXXstrcpy(mnt_type, tmp+strlen("fs="));
+                    tmp = strchr(mnt_type, ',');
+                    if ( tmp ) {
+                        *tmp = '\0';
+                    }
+                }
+		if(!strcmp(mnt_type, "cdfss"))
+		    XXXstrcpy(mnt_type, MNTTYPE_CDROM);
+            }
+
             tmp = strstr(mntent->mnt_opts, "loop=");
             if ( tmp ) {
-                strcpy(mntdev, tmp+strlen("loop="));
+                XXXstrcpy(mntdev, tmp+strlen("loop="));
                 tmp = strchr(mntdev, ',');
                 if ( tmp ) {
                     *tmp = '\0';
                 }
             }
+#undef XXXstrcpy
 
             if( strncmp(mntdev, "/dev", 4) ||
                 realpath(mntdev, mntdevpath) == NULL ) {

cu
Ludwig

-- 
(o_  Ludwig.Nussel at gmx.de
//\  PGP Key ID: FF8135CE
V_/_ ICQ:        52166811


From ludwig.nussel at gmx.de  Sat Apr 17 15:42:34 2004
From: ludwig.nussel at gmx.de (Ludwig Nussel)
Date: Sat, 17 Apr 2004 21:42:34 +0200
Subject: [PATCH] freeing the product info
Message-ID: <20040417194234.GB1862@defiant.wachendorf.lan>

Hi,

Handling of info->product is broken. It just can't work if you assign
something to info->product, free it and then try to access it. That
happens for example when installers have post install scripts.

Since info->product ist created in create_install(), I propose to
free it in delete_install() instead of generate_uninstall(). I
tested the component install and uninstall with Rune HoV, worked
fine with the patch.


Index: loki_setup/install.c
===================================================================
--- loki_setup.orig/install.c	2004-04-17 19:31:46.553443982 +0200
+++ loki_setup/install.c	2004-04-17 19:48:33.033714876 +0200
@@ -1410,6 +1410,12 @@
 		free(comp->message);
         free(comp);
     }
+	if(info->product)
+	{
+		loki_closeproduct(info->product);
+		info->product = NULL;
+	}
+
 	delete_cdrom_install(info);
     if ( info->lookup ) {
         close_lookup(info->lookup);
@@ -1763,10 +1769,8 @@
     product_option_t *option;
     struct component_elem *comp;
 
-    if ( info->component ) { /* Component install, the product has already been opened */
-        product = info->product;
-        component = info->component;
-    } else {
+	if(!info->product) {
+		log_debug("create new product");
         /* Try to open the product first in case it was installed previously */
         product = loki_openproduct(info->name);
         if ( ! product ) {
@@ -1779,6 +1783,9 @@
 		info->product = product;
     }
 
+	product = info->product;
+	component = info->component;
+
     if ( product ) {
         char buf[PATH_MAX];
 		struct envvar_elem *var;
@@ -1951,8 +1958,6 @@
 
         snprintf(buf, sizeof(buf), "setup.data/bin/%s/%s/uninstall", detect_os(), detect_arch());
         loki_upgrade_uninstall(product, buf, "setup.data/locale");
-		/* We must call the following in all cases - component installs even, as we have to save the changes */
-		loki_closeproduct(product);
     } else {
 		log_fatal(_("Could not create install log"),
 				  detect_home(), info->name);

cu
Ludwig

-- 
(o_  Ludwig.Nussel at gmx.de
//\  PGP Key ID: FF8135CE
V_/_ ICQ:        52166811


From megastep at megastep.org  Sat Apr 17 20:11:18 2004
From: megastep at megastep.org (=?ISO-8859-1?Q?St=E9phane?= Peter)
Date: Sat, 17 Apr 2004 17:11:18 -0700
Subject: [lokisetup] [PATCH] subfs
In-Reply-To: <20040417193442.GA1862@defiant.wachendorf.lan>
References: <20040417193442.GA1862@defiant.wachendorf.lan>
Message-ID: <1082247078.3104.11.camel@megaserv.megastep-la.org>

Cool :) I'm committing this patch as well as the previous one that fixed
the uninstall.c compilatio.

> Hi,
> 
> SUSE 9.1 uses subfs to automatically mount removable media.
> Unfortunately subfs mounted CD-ROMs don't appear in /etc/mtab so
> loki-setup doesn't find them. A subfs line in /etc/fstab looks like
> this:
> 
> /dev/cdrom /media/cdrom subfs fs=cdfss,ro,procuid,nosuid,nodev,exec,iocharset=utf8 0 0
> 
> ut2004 installation works with the following patch:
[...]
> cu
> Ludwig

-- 
Stephane Peter
Sr. Software Engineer
Codehost, Inc.



From megastep at megastep.org  Sat Apr 17 20:20:13 2004
From: megastep at megastep.org (=?ISO-8859-1?Q?St=E9phane?= Peter)
Date: Sat, 17 Apr 2004 17:20:13 -0700
Subject: [lokisetup] [PATCH] freeing the product info
In-Reply-To: <20040417194234.GB1862@defiant.wachendorf.lan>
References: <20040417194234.GB1862@defiant.wachendorf.lan>
Message-ID: <1082247613.3104.23.camel@megaserv.megastep-la.org>

I'm a bit nervous to commit this because the last patch addressing this
problem actually broke component installs for me. I'll try this but I've
been using the current CVS code in a commercial product with
post-install scripts among a bunch of platforms with no problem so far.
So I'm not sure why this wouldn't work for you ?

> Hi,
> 
> Handling of info->product is broken. It just can't work if you assign
> something to info->product, free it and then try to access it. That
> happens for example when installers have post install scripts.
> 
> Since info->product ist created in create_install(), I propose to
> free it in delete_install() instead of generate_uninstall(). I
> tested the component install and uninstall with Rune HoV, worked
> fine with the patch.
> 
> 
> Index: loki_setup/install.c
> ===================================================================
> --- loki_setup.orig/install.c	2004-04-17 19:31:46.553443982 +0200
> +++ loki_setup/install.c	2004-04-17 19:48:33.033714876 +0200
> @@ -1410,6 +1410,12 @@
>  		free(comp->message);
>          free(comp);
>      }
> +	if(info->product)
> +	{
> +		loki_closeproduct(info->product);
> +		info->product = NULL;
> +	}
> +
>  	delete_cdrom_install(info);
>      if ( info->lookup ) {
>          close_lookup(info->lookup);
> @@ -1763,10 +1769,8 @@
>      product_option_t *option;
>      struct component_elem *comp;
>  
> -    if ( info->component ) { /* Component install, the product has already been opened */
> -        product = info->product;
> -        component = info->component;
> -    } else {
> +	if(!info->product) {
> +		log_debug("create new product");
>          /* Try to open the product first in case it was installed previously */
>          product = loki_openproduct(info->name);
>          if ( ! product ) {
> @@ -1779,6 +1783,9 @@
>  		info->product = product;
>      }
>  
> +	product = info->product;
> +	component = info->component;
> +
>      if ( product ) {
>          char buf[PATH_MAX];
>  		struct envvar_elem *var;
> @@ -1951,8 +1958,6 @@
>  
>          snprintf(buf, sizeof(buf), "setup.data/bin/%s/%s/uninstall", detect_os(), detect_arch());
>          loki_upgrade_uninstall(product, buf, "setup.data/locale");
> -		/* We must call the following in all cases - component installs even, as we have to save the changes */
> -		loki_closeproduct(product);
>      } else {
>  		log_fatal(_("Could not create install log"),
>  				  detect_home(), info->name);
> 
> cu
> Ludwig
-- 
Stephane Peter
Sr. Software Engineer
Codehost, Inc.



From ludwig.nussel at gmx.de  Sun Apr 18 05:53:16 2004
From: ludwig.nussel at gmx.de (Ludwig Nussel)
Date: Sun, 18 Apr 2004 11:53:16 +0200
Subject: [lokisetup] [PATCH] freeing the product info
In-Reply-To: <1082247613.3104.23.camel@megaserv.megastep-la.org>
References: <20040417194234.GB1862@defiant.wachendorf.lan> <1082247613.3104.23.camel@megaserv.megastep-la.org>
Message-ID: <20040418095316.GA4628@defiant.wachendorf.lan>

St?phane Peter wrote:
> I'm a bit nervous to commit this because the last patch addressing this
> problem actually broke component installs for me. I'll try this but I've
> been using the current CVS code in a commercial product with
> post-install scripts among a bunch of platforms with no problem so far.

Pure luck. Let me reconstruct the problem. Look at
generate_uninstall, it's called by main.c:411->install.c:1505. In
install.c it does 

  1771          product = loki_openproduct(info->name);
  1772          if ( ! product ) {
  1773              product = loki_create_product(info->name, info->install_path, info->desc,
...
  1779          info->product = product;
...
  1955          loki_closeproduct(product);

which in turn does free(product);

so now info->product points to already freed memory!

next, main.c:412 calls install_postinstall() which in turn calls
run_script() which calls get_optiontags_string() at install.c:1527.
And this one finally does

install.c:1669   for ( comp = loki_getfirst_component(info->product)

which is at setupdb.c:786:

product_component_t *loki_getfirst_component(product_t *product)
{
    return product->components;
}

Say hello to Mr Segfault! I suppose glibc decides to defer actually
freeing info->product so it doesn't crash always. It crashed for me
when I compiled loki-setup on SuSE 8.1 with glibc 2.2 and ran it on
a glibc 2.3 system. You can also use valgrind, it will tell you
about illegal reads on already freed memory.

cu
Ludwig

-- 
(o_  Ludwig.Nussel at gmx.de
//\  PGP Key ID: FF8135CE
V_/_ ICQ:        52166811


From megastep at megastep.org  Thu Apr 29 21:26:46 2004
From: megastep at megastep.org (Stephane Peter)
Date: Thu, 29 Apr 2004 18:26:46 -0700
Subject: Memory leak fixes in setupdb and setup
Message-ID: <71FCAB31-9A45-11D8-B138-000A959DD562@megastep.org>

I just committed a bunch of fixes to memory management in both 
loki_setupdb and loki_setup. Basically the root of this problem is that 
in libxml, xmlGetProp() returns a malloced string, that needs to be 
explicitly deallocated. This just came to my attention recently when 
running valgrind on some of my projects, and I started to clean up 
setup as well.

Unfortunately we really abuse this function all over the code. Setupdb 
should be mostly clean now, but loki_setup is going to be a royal pain 
to fix. I only fixed the most easy functions so far. On the other hand, 
the urge is not too strong since it's an installer and the user usually 
doesn't have time to notice the leaks...

Anyway, if you notice any weirdness in the current CVS snapshot, please 
let me know.

--
St?phane Peter
megastep at megastep.org


From icculus at clutteredmind.org  Fri Apr 30 03:04:27 2004
From: icculus at clutteredmind.org (Ryan C. Gordon)
Date: Fri, 30 Apr 2004 03:04:27 -0400
Subject: [lokisetup] Memory leak fixes in setupdb and setup
In-Reply-To: <71FCAB31-9A45-11D8-B138-000A959DD562@megastep.org>
References: <71FCAB31-9A45-11D8-B138-000A959DD562@megastep.org>
Message-ID: <1083308667.1115.2.camel@localhost>


> Anyway, if you notice any weirdness in the current CVS snapshot, please 
> let me know.

Fixed a crashbug from this in copy.c; sync up to the latest CVS.

--ryan.





From icculus at clutteredmind.org  Fri Apr 30 03:25:35 2004
From: icculus at clutteredmind.org (Ryan C. Gordon)
Date: Fri, 30 Apr 2004 03:25:35 -0400
Subject: [lokisetup] Memory leak fixes in setupdb and setup
In-Reply-To: <71FCAB31-9A45-11D8-B138-000A959DD562@megastep.org>
References: <71FCAB31-9A45-11D8-B138-000A959DD562@megastep.org>
Message-ID: <1083309935.1110.7.camel@localhost>


> Anyway, if you notice any weirdness in the current CVS snapshot, please 
> let me know.

Ugh, just hit another bug in copy.c...look for the call to
add_bin_entry()...we free "symlink", but add_bin_entry stores the
pointer from libxml, so you dereference free()'d memory on the "Type
'%s' to start the game" message.

How much memory does libxml leak in the worst case? A few bytes per XML
tag/attribute? I'm inclined to suggest you back out the change and let
it leak, since this is just going to cause a lot of subtle bugs.

Alternately, these are the sort of things valgrind should find...funny
that valgrind spurred this change in the first place.  :)

--ryan.





From megastep at megastep.org  Fri Apr 30 17:48:11 2004
From: megastep at megastep.org (Stephane Peter)
Date: Fri, 30 Apr 2004 14:48:11 -0700
Subject: [lokisetup] Memory leak fixes in setupdb and setup
In-Reply-To: <1083309935.1110.7.camel@localhost>
References: <71FCAB31-9A45-11D8-B138-000A959DD562@megastep.org> <1083309935.1110.7.camel@localhost>
Message-ID: <13174FEB-9AF0-11D8-B138-000A959DD562@megastep.org>

OK, i have to admit I didn't try out the changes before committing. I 
just spent a couple of days doing the exact same kind of changes to a 
lot of code, but errors slip by...

I agree this is potentially dangerous, but I'm rather inclined myself 
to do things right, at least fix the more obvious leaks. The leaks are 
actually a few bytes per tag, but unfortunately this is often code that 
is called often. Overall, yeah, the leaks are barely noticeable.

I'm gonna have to release an installer soon for my own purposes, so 
I'll run this through valgrind and debug any more problem I find...

On another topic, I also committed changes to detect Gentoo in setupdb. 
:)

On Apr 30, 2004, at 12:25 AM, Ryan C. Gordon wrote:
>> Anyway, if you notice any weirdness in the current CVS snapshot, 
>> please
>> let me know.
>
> Ugh, just hit another bug in copy.c...look for the call to
> add_bin_entry()...we free "symlink", but add_bin_entry stores the
> pointer from libxml, so you dereference free()'d memory on the "Type
> '%s' to start the game" message.
>
> How much memory does libxml leak in the worst case? A few bytes per XML
> tag/attribute? I'm inclined to suggest you back out the change and let
> it leak, since this is just going to cause a lot of subtle bugs.
>
> Alternately, these are the sort of things valgrind should find...funny
> that valgrind spurred this change in the first place.  :)
>
> --ryan.
>
>
>
--
St?phane Peter
megastep at megastep.org