[PATCH] package_ipk: apply umask to control and conffiles

[PATCH] package_ipk: apply umask to control and conffiles

[Andreas Oberritter]

* Explicitly set umask to 022. Otherwise the build system's
  umask leaks into the image.

Signed-off-by: Andreas Oberritter <obi@opendreambox.org>
---
* I tried do_package_ipk[umask] = "022" first, but it didn't work.

 meta/classes/package_ipk.bbclass |    4 ++++
 1 files changed, 4 insertions(+), 0 deletions(-)

diff --git a/meta/classes/package_ipk.bbclass b/meta/classes/package_ipk.bbclass
index 565ef93..e7da07a 100644
--- a/meta/classes/package_ipk.bbclass
+++ b/meta/classes/package_ipk.bbclass
@@ -317,7 +317,9 @@ python do_package_ipk () {
 		controldir = os.path.join(root, 'CONTROL')
 		bb.mkdirhier(controldir)
 		try:
+			mask = os.umask(022)
 			ctrlfile = file(os.path.join(controldir, 'control'), 'w')
+			os.umask(mask)
 		except OSError:
 			bb.utils.unlockfile(lf)
 			raise bb.build.FuncFailed("unable to open control file for writing.")
@@ -410,7 +412,9 @@ python do_package_ipk () {
 		conffiles_str = localdata.getVar("CONFFILES", True)
 		if conffiles_str:
 			try:
+				mask = os.umask(022)
 				conffiles = file(os.path.join(controldir, 'conffiles'), 'w')
+				os.umask(mask)
 			except OSError:
 				bb.utils.unlockfile(lf)
 				raise bb.build.FuncFailed("unable to open conffiles for writing.")
-- 
1.7.5.4

Re: [PATCH] package_ipk: apply umask to control and conffiles

[Mark Hatle]

On 3/9/12 8:15 PM, Andreas Oberritter wrote:
> * Explicitly set umask to 022. Otherwise the build system's
>    umask leaks into the image.

I'm surprised that do_package_ipk[umask] didn't work.  Perhaps its the way it's 
being invoked that is the issue.  (If bitbake doesn't run it, but something else 
does.. then the umask setting doesn't get used.)

As for the change of the umask, the changes appear to be specific to the ipk 
case.  Is this the desired behavior, or could deb and rpm suffer from similar 
issues?  (I'm not familiar enough with opkg to know how it handles umask 
settings during package install/rootfs construction..)

I believe that RPM sets a default umask when it goes through it's package 
installs/rootfs generation.  But does DEB?

--Mark

> Signed-off-by: Andreas Oberritter<obi@opendreambox.org>
> ---
> * I tried do_package_ipk[umask] = "022" first, but it didn't work.
>
>   meta/classes/package_ipk.bbclass |    4 ++++
>   1 files changed, 4 insertions(+), 0 deletions(-)
>
> diff --git a/meta/classes/package_ipk.bbclass b/meta/classes/package_ipk.bbclass
> index 565ef93..e7da07a 100644
> --- a/meta/classes/package_ipk.bbclass
> +++ b/meta/classes/package_ipk.bbclass
> @@ -317,7 +317,9 @@ python do_package_ipk () {
>   		controldir = os.path.join(root, 'CONTROL')
>   		bb.mkdirhier(controldir)
>   		try:
> +			mask = os.umask(022)
>   			ctrlfile = file(os.path.join(controldir, 'control'), 'w')
> +			os.umask(mask)
>   		except OSError:
>   			bb.utils.unlockfile(lf)
>   			raise bb.build.FuncFailed("unable to open control file for writing.")
> @@ -410,7 +412,9 @@ python do_package_ipk () {
>   		conffiles_str = localdata.getVar("CONFFILES", True)
>   		if conffiles_str:
>   			try:
> +				mask = os.umask(022)
>   				conffiles = file(os.path.join(controldir, 'conffiles'), 'w')
> +				os.umask(mask)
>   			except OSError:
>   				bb.utils.unlockfile(lf)
>   				raise bb.build.FuncFailed("unable to open conffiles for writing.")

Re: [PATCH] package_ipk: apply umask to control and conffiles

[Richard Purdie]

On Mon, 2012-03-12 at 10:29 -0500, Mark Hatle wrote:
> On 3/9/12 8:15 PM, Andreas Oberritter wrote:
> > * Explicitly set umask to 022. Otherwise the build system's
> >    umask leaks into the image.
> 
> I'm surprised that do_package_ipk[umask] didn't work.  Perhaps its the way it's 
> being invoked that is the issue.  (If bitbake doesn't run it, but something else 
> does.. then the umask setting doesn't get used.)
> 
> As for the change of the umask, the changes appear to be specific to the ipk 
> case.  Is this the desired behavior, or could deb and rpm suffer from similar 
> issues?  (I'm not familiar enough with opkg to know how it handles umask 
> settings during package install/rootfs construction..)
> 
> I believe that RPM sets a default umask when it goes through it's package 
> installs/rootfs generation.  But does DEB?

I'm also a bit worried about this patch. I'd like to understand why a
task level umask doesn't work. That shouldn't even make any difference
since the permissions/owners/users from install should be getting
used...

Cheers,

Richard

Re: [PATCH] package_ipk: apply umask to control and conffiles

[Andreas Oberritter]

On 12.03.2012 16:53, Richard Purdie wrote:
> On Mon, 2012-03-12 at 10:29 -0500, Mark Hatle wrote:
>> On 3/9/12 8:15 PM, Andreas Oberritter wrote:
>>> * Explicitly set umask to 022. Otherwise the build system's
>>>    umask leaks into the image.
>>
>> I'm surprised that do_package_ipk[umask] didn't work.  Perhaps its the way it's 
>> being invoked that is the issue.  (If bitbake doesn't run it, but something else 
>> does.. then the umask setting doesn't get used.)
>>
>> As for the change of the umask, the changes appear to be specific to the ipk 
>> case.  Is this the desired behavior, or could deb and rpm suffer from similar 
>> issues?  (I'm not familiar enough with opkg to know how it handles umask 
>> settings during package install/rootfs construction..)
>>
>> I believe that RPM sets a default umask when it goes through it's package 
>> installs/rootfs generation.  But does DEB?
> 
> I'm also a bit worried about this patch. I'd like to understand why a
> task level umask doesn't work. That shouldn't even make any difference
> since the permissions/owners/users from install should be getting
> used...

The two per-package files "conffiles" and "control" affected by this
patch don't get installed by recipes. package_ipk.bbclass creates them.

Regards,
Andreas

Re: [PATCH] package_ipk: apply umask to control and conffiles

[Andreas Oberritter]

On 12.03.2012 16:53, Richard Purdie wrote:
> On Mon, 2012-03-12 at 10:29 -0500, Mark Hatle wrote:
>> On 3/9/12 8:15 PM, Andreas Oberritter wrote:
>>> * Explicitly set umask to 022. Otherwise the build system's
>>>    umask leaks into the image.
>>
>> I'm surprised that do_package_ipk[umask] didn't work.  Perhaps its the way it's 
>> being invoked that is the issue.  (If bitbake doesn't run it, but something else 
>> does.. then the umask setting doesn't get used.)
>>
>> As for the change of the umask, the changes appear to be specific to the ipk 
>> case.  Is this the desired behavior, or could deb and rpm suffer from similar 
>> issues?  (I'm not familiar enough with opkg to know how it handles umask 
>> settings during package install/rootfs construction..)
>>
>> I believe that RPM sets a default umask when it goes through it's package 
>> installs/rootfs generation.  But does DEB?
> 
> I'm also a bit worried about this patch. I'd like to understand why a
> task level umask doesn't work. That shouldn't even make any difference
> since the permissions/owners/users from install should be getting
> used...
> 
> Cheers,
> 
> Richard

Richard,

can you please give some advise on how to continue with this issue?

Regards,
Andreas

Re: [PATCH] package_ipk: apply umask to control and conffiles

[Richard Purdie]

On Fri, 2012-03-23 at 21:17 +0100, Andreas Oberritter wrote:
> On 12.03.2012 16:53, Richard Purdie wrote:
> > On Mon, 2012-03-12 at 10:29 -0500, Mark Hatle wrote:
> >> On 3/9/12 8:15 PM, Andreas Oberritter wrote:
> >>> * Explicitly set umask to 022. Otherwise the build system's
> >>>    umask leaks into the image.
> >>
> >> I'm surprised that do_package_ipk[umask] didn't work.  Perhaps its the way it's 
> >> being invoked that is the issue.  (If bitbake doesn't run it, but something else 
> >> does.. then the umask setting doesn't get used.)
> >>
> >> As for the change of the umask, the changes appear to be specific to the ipk 
> >> case.  Is this the desired behavior, or could deb and rpm suffer from similar 
> >> issues?  (I'm not familiar enough with opkg to know how it handles umask 
> >> settings during package install/rootfs construction..)
> >>
> >> I believe that RPM sets a default umask when it goes through it's package 
> >> installs/rootfs generation.  But does DEB?
> > 
> > I'm also a bit worried about this patch. I'd like to understand why a
> > task level umask doesn't work. That shouldn't even make any difference
> > since the permissions/owners/users from install should be getting
> > used...
>
> can you please give some advise on how to continue with this issue?

I understand half the problem now, the files with the issues are ones
created during the package_ipk task. That addresses one of my big
concerns.

The second thing I'd like to understand is why a task level umask
doesn't resolve this. Looking at what you tried, this might be as simple
as a typo:

do_package_ipk[umask] = "022"

when you really want:

do_package_write_ipk[umask] = "022"

If that works, lets set this for deb and rpm too so we're consistent and
I'll merge that patch :)

Cheers,

Richard

Re: [PATCH] package_ipk: apply umask to control and conffiles

[Andreas Oberritter]

On 26.03.2012 13:34, Richard Purdie wrote:
> On Fri, 2012-03-23 at 21:17 +0100, Andreas Oberritter wrote:
>> On 12.03.2012 16:53, Richard Purdie wrote:
>>> On Mon, 2012-03-12 at 10:29 -0500, Mark Hatle wrote:
>>>> On 3/9/12 8:15 PM, Andreas Oberritter wrote:
>>>>> * Explicitly set umask to 022. Otherwise the build system's
>>>>>    umask leaks into the image.
>>>>
>>>> I'm surprised that do_package_ipk[umask] didn't work.  Perhaps its the way it's 
>>>> being invoked that is the issue.  (If bitbake doesn't run it, but something else 
>>>> does.. then the umask setting doesn't get used.)
>>>>
>>>> As for the change of the umask, the changes appear to be specific to the ipk 
>>>> case.  Is this the desired behavior, or could deb and rpm suffer from similar 
>>>> issues?  (I'm not familiar enough with opkg to know how it handles umask 
>>>> settings during package install/rootfs construction..)
>>>>
>>>> I believe that RPM sets a default umask when it goes through it's package 
>>>> installs/rootfs generation.  But does DEB?
>>>
>>> I'm also a bit worried about this patch. I'd like to understand why a
>>> task level umask doesn't work. That shouldn't even make any difference
>>> since the permissions/owners/users from install should be getting
>>> used...
>>
>> can you please give some advise on how to continue with this issue?
> 
> I understand half the problem now, the files with the issues are ones
> created during the package_ipk task. That addresses one of my big
> concerns.
> 
> The second thing I'd like to understand is why a task level umask
> doesn't resolve this. Looking at what you tried, this might be as simple
> as a typo:
> 
> do_package_ipk[umask] = "022"
> 
> when you really want:
> 
> do_package_write_ipk[umask] = "022"

Richard,

thank you, that did it. It wasn't a typo, but lack of understanding of 
the "magic" behind it, as I tried to apply the umask to the 
do_package_ipk function instead of the do_package_write_ipk *task*.

Please see below for an updated patch.

Regards,
Andreas

> If that works, lets set this for deb and rpm too so we're consistent and
> I'll merge that patch :)

From f9e4707d7a619e29530ac144ae1aba9d9e406884 Mon Sep 17 00:00:00 2001
From: Andreas Oberritter <obi@opendreambox.org>
Date: Sat, 10 Mar 2012 03:11:57 +0100
Subject: [PATCH] package_{deb,ipk,rpm}: apply umask to files generated in
 do_package_write_{deb,ipk,rpm}

* Explicitly set umask to 022. Otherwise the build system's
  umask may leak into the image.

Signed-off-by: Andreas Oberritter <obi@opendreambox.org>
---
 meta/classes/package_deb.bbclass |    1 +
 meta/classes/package_ipk.bbclass |    1 +
 meta/classes/package_rpm.bbclass |    1 +
 3 files changed, 3 insertions(+), 0 deletions(-)

diff --git a/meta/classes/package_deb.bbclass b/meta/classes/package_deb.bbclass
index 1f7ec9c..9880258 100644
--- a/meta/classes/package_deb.bbclass
+++ b/meta/classes/package_deb.bbclass
@@ -421,6 +421,7 @@ python do_package_write_deb () {
 	bb.build.exec_func("do_package_deb", d)
 }
 do_package_write_deb[dirs] = "${PKGWRITEDIRDEB}"
+do_package_write_deb[umask] = "022"
 addtask package_write_deb before do_package_write after do_package
 
 
diff --git a/meta/classes/package_ipk.bbclass b/meta/classes/package_ipk.bbclass
index 565ef93..e4a217b 100644
--- a/meta/classes/package_ipk.bbclass
+++ b/meta/classes/package_ipk.bbclass
@@ -453,6 +453,7 @@ python do_package_write_ipk () {
 	bb.build.exec_func("do_package_ipk", d)
 }
 do_package_write_ipk[dirs] = "${PKGWRITEDIRIPK}"
+do_package_write_ipk[umask] = "022"
 addtask package_write_ipk before do_package_write after do_package
 
 PACKAGEINDEXES += "package_update_index_ipk;"
diff --git a/meta/classes/package_rpm.bbclass b/meta/classes/package_rpm.bbclass
index e83fc55..daa9301 100644
--- a/meta/classes/package_rpm.bbclass
+++ b/meta/classes/package_rpm.bbclass
@@ -1056,6 +1056,7 @@ python do_package_write_rpm () {
 }
 
 do_package_write_rpm[dirs] = "${PKGWRITEDIRRPM}"
+do_package_write_rpm[umask] = "022"
 addtask package_write_rpm before do_package_write after do_package
 
 PACKAGEINDEXES += "package_update_index_rpm; createrepo ${DEPLOY_DIR_RPM};"
-- 
1.7.5.4

Re: [PATCH] package_ipk: apply umask to control and conffiles

[Saul Wold]

On 03/26/2012 08:47 AM, Andreas Oberritter wrote:
> On 26.03.2012 13:34, Richard Purdie wrote:
>> On Fri, 2012-03-23 at 21:17 +0100, Andreas Oberritter wrote:
>>> On 12.03.2012 16:53, Richard Purdie wrote:
>>>> On Mon, 2012-03-12 at 10:29 -0500, Mark Hatle wrote:
>>>>> On 3/9/12 8:15 PM, Andreas Oberritter wrote:
>>>>>> * Explicitly set umask to 022. Otherwise the build system's
>>>>>>     umask leaks into the image.
>>>>>
>>>>> I'm surprised that do_package_ipk[umask] didn't work.  Perhaps its the way it's
>>>>> being invoked that is the issue.  (If bitbake doesn't run it, but something else
>>>>> does.. then the umask setting doesn't get used.)
>>>>>
>>>>> As for the change of the umask, the changes appear to be specific to the ipk
>>>>> case.  Is this the desired behavior, or could deb and rpm suffer from similar
>>>>> issues?  (I'm not familiar enough with opkg to know how it handles umask
>>>>> settings during package install/rootfs construction..)
>>>>>
>>>>> I believe that RPM sets a default umask when it goes through it's package
>>>>> installs/rootfs generation.  But does DEB?
>>>>
>>>> I'm also a bit worried about this patch. I'd like to understand why a
>>>> task level umask doesn't work. That shouldn't even make any difference
>>>> since the permissions/owners/users from install should be getting
>>>> used...
>>>
>>> can you please give some advise on how to continue with this issue?
>>
>> I understand half the problem now, the files with the issues are ones
>> created during the package_ipk task. That addresses one of my big
>> concerns.
>>
>> The second thing I'd like to understand is why a task level umask
>> doesn't resolve this. Looking at what you tried, this might be as simple
>> as a typo:
>>
>> do_package_ipk[umask] = "022"
>>
>> when you really want:
>>
>> do_package_write_ipk[umask] = "022"
>
> Richard,
>
> thank you, that did it. It wasn't a typo, but lack of understanding of
> the "magic" behind it, as I tried to apply the umask to the
> do_package_ipk function instead of the do_package_write_ipk *task*.
>
> Please see below for an updated patch.
>
> Regards,
> Andreas
>
>> If that works, lets set this for deb and rpm too so we're consistent and
>> I'll merge that patch :)
>
>> From f9e4707d7a619e29530ac144ae1aba9d9e406884 Mon Sep 17 00:00:00 2001
> From: Andreas Oberritter<obi@opendreambox.org>
> Date: Sat, 10 Mar 2012 03:11:57 +0100
> Subject: [PATCH] package_{deb,ipk,rpm}: apply umask to files generated in
>   do_package_write_{deb,ipk,rpm}
>
> * Explicitly set umask to 022. Otherwise the build system's
>    umask may leak into the image.
>
> Signed-off-by: Andreas Oberritter<obi@opendreambox.org>
> ---
>   meta/classes/package_deb.bbclass |    1 +
>   meta/classes/package_ipk.bbclass |    1 +
>   meta/classes/package_rpm.bbclass |    1 +
>   3 files changed, 3 insertions(+), 0 deletions(-)
>
> diff --git a/meta/classes/package_deb.bbclass b/meta/classes/package_deb.bbclass
> index 1f7ec9c..9880258 100644
> --- a/meta/classes/package_deb.bbclass
> +++ b/meta/classes/package_deb.bbclass
> @@ -421,6 +421,7 @@ python do_package_write_deb () {
>   	bb.build.exec_func("do_package_deb", d)
>   }
>   do_package_write_deb[dirs] = "${PKGWRITEDIRDEB}"
> +do_package_write_deb[umask] = "022"
>   addtask package_write_deb before do_package_write after do_package
>
>
> diff --git a/meta/classes/package_ipk.bbclass b/meta/classes/package_ipk.bbclass
> index 565ef93..e4a217b 100644
> --- a/meta/classes/package_ipk.bbclass
> +++ b/meta/classes/package_ipk.bbclass
> @@ -453,6 +453,7 @@ python do_package_write_ipk () {
>   	bb.build.exec_func("do_package_ipk", d)
>   }
>   do_package_write_ipk[dirs] = "${PKGWRITEDIRIPK}"
> +do_package_write_ipk[umask] = "022"
>   addtask package_write_ipk before do_package_write after do_package
>
>   PACKAGEINDEXES += "package_update_index_ipk;"
> diff --git a/meta/classes/package_rpm.bbclass b/meta/classes/package_rpm.bbclass
> index e83fc55..daa9301 100644
> --- a/meta/classes/package_rpm.bbclass
> +++ b/meta/classes/package_rpm.bbclass
> @@ -1056,6 +1056,7 @@ python do_package_write_rpm () {
>   }
>
>   do_package_write_rpm[dirs] = "${PKGWRITEDIRRPM}"
> +do_package_write_rpm[umask] = "022"
>   addtask package_write_rpm before do_package_write after do_package
>
>   PACKAGEINDEXES += "package_update_index_rpm; createrepo ${DEPLOY_DIR_RPM};"

Merged this patch to OE-Core

Thanks
	Sau!