Re: [PATCH 1/1] sstate.bbclass: update the timestamps after install

[Robert Yang <liezhi.yang@windriver.com>]

On 09/04/2014 04:53 PM, Richard Purdie wrote:
> On Thu, 2014-09-04 at 16:49 +0800, Robert Yang wrote:
>>
>> On 09/04/2014 04:27 PM, Richard Purdie wrote:
>>> On Thu, 2014-09-04 at 00:05 -0700, Robert Yang wrote:
>>>> Update the sstate file's timestamps after it is installed, it will be
>>>> very useful for removing the old sstate file, especially, it's not easy
>>>> to remove when use the shared SSTATE_DIR, we can easily remove them with
>>>> this change, for example:
>>>>
>>>> $ find state-cache -type f -ctime +10 -exec rm -f {} \;
>>>>
>>>> Will remove the sstate file which isn't used by recent 10 days.
>>>>
>>>> We can use the -atime, but it is not always available, for example,
>>>> when mounted with "-o noatime".
>>>>
>>>> The touch is a very light weight action, and the
>>>> scripts/sstate-cache-management.sh also requires this.
>>>>
>>>> Signed-off-by: Robert Yang <liezhi.yang@windriver.com>
>>>> ---
>>>>    meta/classes/sstate.bbclass |    1 +
>>>>    1 file changed, 1 insertion(+)
>>>>
>>>> diff --git a/meta/classes/sstate.bbclass b/meta/classes/sstate.bbclass
>>>> index ead829e..885912d 100644
>>>> --- a/meta/classes/sstate.bbclass
>>>> +++ b/meta/classes/sstate.bbclass
>>>> @@ -618,6 +618,7 @@ sstate_unpack_package () {
>>>>    	mkdir -p ${SSTATE_INSTDIR}
>>>>    	cd ${SSTATE_INSTDIR}
>>>>    	tar -xmvzf ${SSTATE_PKG}
>>>> +	touch --no-dereference ${SSTATE_PKG}
>>>>    }
>>>>
>>>>    BB_HASHCHECK_FUNCTION = "sstate_checkhashes"
>>>
>>> At the very least we need to consider read only files here...
>>
>> Hello, did you mean the SSTATE_MIRRORS ? I thought that I had
>> considered it since I used the "touch --no-dereference",
>> make a clear check is reasonable if I understand correctly,
>> so updated the code in the repo: (The --no-dereference is not
>> need any more since the "test -w" follows symlink).
>
> We need the --no-dereferece but this could also be a readonly directory

Under more thoughts, yes, we need the --no-dereferece otherwise the
touch may update the file in the SSTATE_MIRRORS, and it's safe to touch
a symlink itself (though "test -w" follows symlink) since the permissions
of symbolic links are never used.

For readonly_dir/ssfile, we can touch the ssfile if "test -w ssfile",
it doesn't care about the dir's write permission. Or do we have to
check the permission of "dirname ${SSTATE_PKG}" and ${SSTATE_DIR} ?
I don't think that we need unless it causes errors.

For the ISO, the "test -w {SSTATE_PKG}" would fail so that no touch,
here is the updated code, also in the repo:

diff --git a/meta/classes/sstate.bbclass b/meta/classes/sstate.bbclass
index ead829e..78766c0 100644
--- a/meta/classes/sstate.bbclass
+++ b/meta/classes/sstate.bbclass
@@ -618,6 +618,8 @@ sstate_unpack_package () {
         mkdir -p ${SSTATE_INSTDIR}
         cd ${SSTATE_INSTDIR}
         tar -xmvzf ${SSTATE_PKG}
+       # Use "! -w ||" to return true for read only files
+       [ ! -w ${SSTATE_PKG} ] || touch ${SSTATE_PKG}
  }

// Robert

> or set of readonly files (think an ISO shipped with WR Linux or locked
> sstate cache).
>
> Cheers,
>
> Richard
>
>
>