Closed Bug 708793 Opened 10 years ago Closed 10 years ago

provide way to override platform name in POST

Categories

(Webtools Graveyard :: Graph Server, defect)

defect
Not set
normal

Tracking

(Not tracked)

RESOLVED FIXED

People

(Reporter: rhelmer, Assigned: jmaher)

References

Details

Attachments

(2 files, 1 obsolete file)

Right now graphserver infers the platform name from the machine name. Provide a way to override this (see bug 708346 comment 3 for more info).
Assignee: nobody → rhelmer
hmm, it seems as though this is not a simple change;  maybe a solution is to put a full set of machine names in for the new platform into the database:

insert into machines values (NULL,20,0,1.0,"tegra-001",1,unix_timestamp());

*add a .n and change the os_id*
insert into machines values (NULL,21,0,1.0,"tegra-001.n",1,unix_timestamp());

Then upon posting we can determine if there is a platform override and switch on that.  

Otherwise, I see this as being a lot more surgery since our os_id is tied so closely to the machine names.
(In reply to Joel Maher (:jmaher) from comment #1)
> hmm, it seems as though this is not a simple change;  maybe a solution is to
> put a full set of machine names in for the new platform into the database:
> 
> insert into machines values (NULL,20,0,1.0,"tegra-001",1,unix_timestamp());
> 
> *add a .n and change the os_id*
> insert into machines values (NULL,21,0,1.0,"tegra-001.n",1,unix_timestamp());
> 
> Then upon posting we can determine if there is a platform override and
> switch on that.  
> 
> Otherwise, I see this as being a lot more surgery since our os_id is tied so
> closely to the machine names.

Would it be worth faking the machine name, at least as an interim solution?
that is probably doable.  Let me work with bear to figure out a reasonable solution for this.
(In reply to Robert Helmer [:rhelmer] from comment #2)
> (In reply to Joel Maher (:jmaher) from comment #1)
> > hmm, it seems as though this is not a simple change;  maybe a solution is to
> > put a full set of machine names in for the new platform into the database:
> > 
> > insert into machines values (NULL,20,0,1.0,"tegra-001",1,unix_timestamp());
> > 
> > *add a .n and change the os_id*
> > insert into machines values (NULL,21,0,1.0,"tegra-001.n",1,unix_timestamp());
> > 
> > Then upon posting we can determine if there is a platform override and
> > switch on that.  
> > 
> > Otherwise, I see this as being a lot more surgery since our os_id is tied so
> > closely to the machine names.
> 
> Would it be worth faking the machine name, at least as an interim solution?

Oh re-reading, that's basically what you're proposing right? Do we even need to do a platform override, or would Talos just post with "tegra-001.n" as the machine name?
(In reply to Joel Maher (:jmaher) from comment #3)
> that is probably doable.  Let me work with bear to figure out a reasonable
> solution for this.

OK cool, thanks! Let me know if you need anything.

Longer-term, we should change the assumption in graphserver that a machine_id will only ever test one platform.
yeah, that is what I suggested as a work around.  we just need to add definitions for the other test type.
patch to add a android native machine type as well as copying the tegra definitions into tegra.n names.
Attachment #582029 - Flags: review?(bear)
this adds a --nativeUI cli flag to remotePerfConfigurator.py so we can s/title/title.n/ to resolve this bug.

Pretty straightforward patch!
Attachment #582037 - Flags: review?(jhammel)
Attachment #582037 - Flags: feedback?(bear)
Comment on attachment 582037 [details] [diff] [review]
[talos] add a --nativeUI flag so we can tweak the machine name (1.0)

Looks fine.  I'm a bit confused about the '.n' suffix for self.title; maybe worth a comment?
Attachment #582037 - Flags: review?(jhammel) → review+
updated with a comment:
        # For NativeUI Fennec, we are working around bug 708793 and uploading a 
        # unique machine name (defined via title) with a .n.  Currently a machine name
        # is a 1:1 mapping with the OS+hardware
Attachment #582037 - Attachment is obsolete: true
Attachment #582037 - Flags: feedback?(bear)
Attachment #582046 - Flags: review+
Attachment #582046 - Flags: feedback?(bear)
Attachment #582029 - Flags: review?(bear) → review+
Attachment #582046 - Flags: feedback?(bear) → feedback+
Depends on: 711365
Depends on: 711367
(In reply to Joel Maher (:jmaher) from comment #12)
> landed on graphs:
> http://hg.mozilla.org/graphs/rev/c3dc35aaef73

Thanks for taking this, do we need to do anything further here? Is it deployed etc?
Assignee: rhelmer → jmaher
Status: NEW → RESOLVED
Closed: 10 years ago
Resolution: --- → FIXED
this is deployed and live.  I think for now this will work great.  Maybe as future changes get in we could fix this in a more generic method.
Blocks: 952213
Product: Webtools → Webtools Graveyard
You need to log in before you can comment on or make changes to this bug.