Processed: Re: Bug#843701: jessie-pu: package boinc/7.4.23+dfsg-1

[Debian Bug Tracking System]

Processing control commands:

> tag -1 pending
Bug #843701 [release.debian.org] jessie-pu: package boinc/7.4.23+dfsg-1
Added tag(s) pending.

-- 
843701: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=843701
Debian Bug Tracking System
Contact ow...@bugs.debian.org with problems

Bug#843701: jessie-pu: package boinc/7.4.23+dfsg-1

[Cyril Brulebois]

Control: tag -1 pending

Gianfranco Costamagna  (2017-06-30):
> >Looks better to me, thanks. Feel free to upload.
> 
> thanks!

Now flagged for acceptance, thanks.


KiBi.


signature.asc
Description: Digital signature

Bug#843701: jessie-pu: package boinc/7.4.23+dfsg-1

[Gianfranco Costamagna]

Hello Cyril,

>Looks better to me, thanks. Feel free to upload.

thanks!
>Also, please use reply-all on the BTS.


ack

G.

Processed: Re: Bug#843701: jessie-pu: package boinc/7.4.23+dfsg-1

[Debian Bug Tracking System]

Processing control commands:

> tag -1 confirmed
Bug #843701 [release.debian.org] jessie-pu: package boinc/7.4.23+dfsg-1
Added tag(s) confirmed.

-- 
843701: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=843701
Debian Bug Tracking System
Contact ow...@bugs.debian.org with problems

Bug#843701: jessie-pu: package boinc/7.4.23+dfsg-1

[Cyril Brulebois]

Control: tag -1 confirmed

Gianfranco Costamagna  (2017-06-27):
> >I'm interested in seeing an updated debdiff with a better wording for
> >the xhost issue. The proposed one suggests a syntax error but says
> >nothing about the permission issues which need a fix.
> 
> ok, fair enough, updated
> 
> >Similarly, the OOM_ADJ handling could be more descriptive, something
> >like “Try both oom_score_adj and oom_adj when adjusting the OOM score
> >(Closes: #843663).”?
> 
> this seems really better and more descriptive, indeed.
> 
> Updated debdiff attached

Looks better to me, thanks. Feel free to upload.

Also, please use reply-all on the BTS.


KiBi.


signature.asc
Description: Digital signature

Processed: Re: Bug#843701: jessie-pu: package boinc/7.4.23+dfsg-1

[Debian Bug Tracking System]

Processing control commands:

> tags -1 -moreinfo
Bug #843701 [release.debian.org] jessie-pu: package boinc/7.4.23+dfsg-1
Removed tag(s) moreinfo.

-- 
843701: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=843701
Debian Bug Tracking System
Contact ow...@bugs.debian.org with problems

Bug#843701: jessie-pu: package boinc/7.4.23+dfsg-1

[Gianfranco Costamagna]

control: tags -1 -moreinfo


>I'm interested in seeing an updated debdiff with a better wording for
>the xhost issue. The proposed one suggests a syntax error but says
>nothing about the permission issues which need a fix.

ok, fair enough, updated

>Similarly, the OOM_ADJ handling could be more descriptive, something
>like “Try both oom_score_adj and oom_adj when adjusting the OOM score
>(Closes: #843663).”?

this seems really better and more descriptive, indeed.

Updated debdiff attached

thanks

G.
diff -Nru boinc-7.4.23+dfsg/debian/boinc-client.init 
boinc-7.4.23+dfsg/debian/boinc-client.init
--- boinc-7.4.23+dfsg/debian/boinc-client.init  2014-10-17 17:10:09.0 
+0200
+++ boinc-7.4.23+dfsg/debian/boinc-client.init  2016-11-08 21:53:59.0 
+0100
@@ -29,6 +29,7 @@
 BOINC_DIR=/var/lib/boinc-client
 BOINC_CLIENT=/usr/bin/boinc
 BOINC_OOM_ADJ=15
+BOINC_OOM_SCORE_ADJ=1000
 
 #VALGRIND_OPTIONS="-v --log-file=/tmp/valgrind_boinc.log "
 VALGRIND_OPTIONS=""
@@ -106,7 +107,7 @@
   else
 if [ -n "$DISPLAY" -a -x /usr/bin/xhost ]; then
# grant the boinc client to perform GPU computing
-   xhost local:boinc || echo -n "xhost error ignored, GPU computing may 
not be possible"
+   xhost +si:localuser:$BOINC_USER || echo -n "xhost error ignored, GPU 
computing may not be possible"
 fi
 if [ -n "$VALGRIND_OPTIONS" ]; then
   start-stop-daemon --start --quiet --background --pidfile $PIDFILE \
@@ -206,10 +207,13 @@
   fi
 fi
 for BPID in ${pid} ${children}; do
-  if [ -w /proc/${BPID}/oom_adj ]; then
-echo ${BOINC_OOM_AD} > /proc/${BPID}/oom_adj 2>/dev/null || true
+  # Fallback to old oom_adj if oom_score_adj doesn't exist
+  if [ -w /proc/${BPID}/oom_score_adj ]; then
+echo ${BOINC_OOM_SCORE_ADJ} > /proc/${BPID}/oom_score_adj 2>/dev/null 
|| true
+  elif [ -w /proc/${BPID}/oom_adj ]; then
+echo ${BOINC_OOM_ADJ} > /proc/${BPID}/oom_adj 2>/dev/null || true
   else
-echo "Could not write to /proc/${BPID}/oom_adj"
+echo "Could not adjust oom_score of task"
   fi
 done
   fi
diff -Nru boinc-7.4.23+dfsg/debian/changelog boinc-7.4.23+dfsg/debian/changelog
--- boinc-7.4.23+dfsg/debian/changelog  2014-10-17 17:19:50.0 +0200
+++ boinc-7.4.23+dfsg/debian/changelog  2016-11-08 21:53:59.0 +0100
@@ -1,3 +1,16 @@
+boinc (7.4.23+dfsg-1+deb8u1) jessie; urgency=medium
+
+  [ Tom Downes ]
+  * Try both oom_score_adj and oom_adj when adjusting the OOM score
+(Closes: #843663).
+
+  [ Mike Brennan  ]
+  * Fix xhost syntax. (Closes: #841665)
+- the xhost permissions syntax requires a "localuser" keyword for locally
+  specified users.
+
+ -- Gianfranco Costamagna   Tue, 08 Nov 2016 
21:53:59 +0100
+
 boinc (7.4.23+dfsg-1) unstable; urgency=medium
 
   * New upstream release candidate.


signature.asc
Description: OpenPGP digital signature

Bug#843701: jessie-pu: package boinc/7.4.23+dfsg-1

[Cyril Brulebois]

Control: tag -1 moreinfo

Hi,

Gianfranco Costamagna  (2016-12-17):
> Hi,
> >Your mail client mangled the diff.
> 
> 
> 
> sorry for that
> 
> > the diff is simple:> +  [ Tom Downes ]
> > +  * Fix OOM_ADJ handling with a backportable approach
> > +(Closes: #843663)
> > 
> > ^^ a typo in a variable name was preventing OOM_ADJ from being correctly 
> > set in the init script
> > 
> 
> What's the impact of that bug?
> 
> when kernel is OOM, boinc tasks should be killed before other tasks.
> boinc is something that shouldn't impact the rest of the system
> (voluntary computing), so it runs with lower nice level, and should
> be killed before other programs in case the system gets out of memory.
> 
> this typo was preventing the second OOM handling to be correctly set, so 
> people might have got
> some other program killed instead of a boinc task.
> 
> >How can that possibly work?  The init script doesn't have an X>display...
> 
> this is a known problem/issue, usually people can do GPU computing with a 
> reload of the boinc-client
> daemon, in this case the init system picks the X server up.
> (this is how things should work, I'm clueless about such stuff and I avoid 
> touching it when it
> "works")
> 
> thanks for the review,

I'm interested in seeing an updated debdiff with a better wording for
the xhost issue. The proposed one suggests a syntax error but says
nothing about the permission issues which need a fix.

Similarly, the OOM_ADJ handling could be more descriptive, something
like “Try both oom_score_adj and oom_adj when adjusting the OOM score
(Closes: #843663).”?


KiBi.


signature.asc
Description: Digital signature

Processed: Re: Bug#843701: jessie-pu: package boinc/7.4.23+dfsg-1

[Debian Bug Tracking System]

Processing control commands:

> tag -1 moreinfo
Bug #843701 [release.debian.org] jessie-pu: package boinc/7.4.23+dfsg-1
Added tag(s) moreinfo.

-- 
843701: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=843701
Debian Bug Tracking System
Contact ow...@bugs.debian.org with problems

Bug#843701: jessie-pu: package boinc/7.4.23+dfsg-1

[Gianfranco Costamagna]

control: tags -1 -moreinfo




Hi,
>Your mail client mangled the diff.



sorry for that

> the diff is simple:> +  [ Tom Downes ]
> +  * Fix OOM_ADJ handling with a backportable approach
> +(Closes: #843663)
> 
> ^^ a typo in a variable name was preventing OOM_ADJ from being correctly set 
> in the init script
> 

What's the impact of that bug?

when kernel is OOM, boinc tasks should be killed before other tasks.
boinc is something that shouldn't impact the rest of the system
(voluntary computing), so it runs with lower nice level, and should
be killed before other programs in case the system gets out of memory.

this typo was preventing the second OOM handling to be correctly set, so people 
might have got
some other program killed instead of a boinc task.

>How can that possibly work?  The init script doesn't have an X>display...

this is a known problem/issue, usually people can do GPU computing with a 
reload of the boinc-client
daemon, in this case the init system picks the X server up.
(this is how things should work, I'm clueless about such stuff and I avoid 
touching it when it
"works")

thanks for the review,

G.

Processed: Re: Bug#843701: jessie-pu: package boinc/7.4.23+dfsg-1

[Debian Bug Tracking System]

Processing control commands:

> tags -1 -moreinfo
Bug #843701 [release.debian.org] jessie-pu: package boinc/7.4.23+dfsg-1
Removed tag(s) moreinfo.

-- 
843701: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=843701
Debian Bug Tracking System
Contact ow...@bugs.debian.org with problems

Bug#843701: jessie-pu: package boinc/7.4.23+dfsg-1

[Julien Cristau]

Control: tag -1 moreinfo

On Tue, Nov  8, 2016 at 21:12:32 +, Gianfranco Costamagna wrote:

> Package: release.debian.org
> User: release.debian@packages.debian.org
> Usertags: pu
> Tags: jessie
> Severity: normal
> 
> Boinc has a functionality problem in stable, the fix is already backported 
> and I did upload
> a new version in unstable a few hours ago
> (just to make it backportable with older kernels)
> 
Your mail client mangled the diff.

> the diff is simple:
> +  [ Tom Downes ]
> +  * Fix OOM_ADJ handling with a backportable approach
> +(Closes: #843663)
> 
> ^^ a typo in a variable name was preventing OOM_ADJ from being correctly set 
> in the init script
> 
What's the impact of that bug?

> 
> +  [ Mike Brennan  ]
> +  * Fix xhost syntax. (Closes: #841665)
> ^^ this is a potential security issue that still affects stable, so I would 
> like to also address it.
> 
How can that possibly work?  The init script doesn't have an X
display...

Cheers,
Julien

Processed: Re: Bug#843701: jessie-pu: package boinc/7.4.23+dfsg-1

[Debian Bug Tracking System]

Processing control commands:

> tag -1 moreinfo
Bug #843701 [release.debian.org] jessie-pu: package boinc/7.4.23+dfsg-1
Added tag(s) moreinfo.

-- 
843701: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=843701
Debian Bug Tracking System
Contact ow...@bugs.debian.org with problems

Bug#843701: jessie-pu: package boinc/7.4.23+dfsg-1

[Gianfranco Costamagna]

Package: release.debian.org
User: release.debian@packages.debian.org
Usertags: pu
Tags: jessie
Severity: normal

Boinc has a functionality problem in stable, the fix is already backported and 
I did upload
a new version in unstable a few hours ago
(just to make it backportable with older kernels)

the diff is simple:
+  [ Tom Downes ]
+  * Fix OOM_ADJ handling with a backportable approach
+(Closes: #843663)

^^ a typo in a variable name was preventing OOM_ADJ from being correctly set in 
the init script


+  [ Mike Brennan  ]
+  * Fix xhost syntax. (Closes: #841665)
^^ this is a potential security issue that still affects stable, so I would 
like to also address it.

diff -Nru boinc-7.4.23+dfsg/debian/boinc-client.init 
boinc-7.4.23+dfsg/debian/boinc-client.init
--- boinc-7.4.23+dfsg/debian/boinc-client.init  2014-10-17 17:10:09.0 
+0200
+++ boinc-7.4.23+dfsg/debian/boinc-client.init  2016-11-08 19:39:13.0 
+0100
@@ -29,6 +29,7 @@
BOINC_DIR=/var/lib/boinc-client
BOINC_CLIENT=/usr/bin/boinc
BOINC_OOM_ADJ=15
+BOINC_OOM_SCORE_ADJ=1000

#VALGRIND_OPTIONS="-v --log-file=/tmp/valgrind_boinc.log "
VALGRIND_OPTIONS=""
@@ -106,7 +107,7 @@
else
if [ -n "$DISPLAY" -a -x /usr/bin/xhost ]; then
# grant the boinc client to perform GPU computing
-   xhost local:boinc || echo -n "xhost error ignored, GPU computing may 
not be possible"
+   xhost +si:localuser:$BOINC_USER || echo -n "xhost error ignored, GPU 
computing may not be possible"
fi
if [ -n "$VALGRIND_OPTIONS" ]; then
start-stop-daemon --start --quiet --background --pidfile $PIDFILE \
@@ -206,10 +207,13 @@
fi
fi
for BPID in ${pid} ${children}; do
-  if [ -w /proc/${BPID}/oom_adj ]; then
-echo ${BOINC_OOM_AD} > /proc/${BPID}/oom_adj 2>/dev/null || true
+  # Fallback to old oom_adj if oom_score_adj doesn't exist
+  if [ -w /proc/${BPID}/oom_score_adj ]; then
+echo ${BOINC_OOM_SCORE_ADJ} > /proc/${BPID}/oom_score_adj 2>/dev/null 
|| true
+  elif [ -w /proc/${BPID}/oom_adj ]; then
+echo ${BOINC_OOM_ADJ} > /proc/${BPID}/oom_adj 2>/dev/null || true
else
-echo "Could not write to /proc/${BPID}/oom_adj"
+echo "Could not adjust oom_score of task"
fi
done
fi


thanks,

Gianfranco


debdiff
Description: Binary data