84122 - Need ability to have failover URLs in mac installer

Status: VERIFIED FIXED

(SeaMonkey :: Installer, defect, P2)

Description

[Samir Gehani]

The [General] section in the mac installer's config.ini has one global URL but
some vendors using mozilla need the ability to failover.

Comment 1

[Samir Gehani]

Transferring keyword, priority, and target milestone decorations from bugscape
5465 that depends on this bug.

Comment 2

[Samir Gehani]

Affects branch and trunk.  Will attach patch for branch only (the codebases are
currently identical).

Comment 3

[Samir Gehani]

Attachment: Patch to support multiple global failover URLs.

Comment 4

[Samir Gehani]

ssu, please r.
alecf, please sr.

Thanks.

Comment 5

[selmer (gone)]

Is this really a 091 stopper?

Comment 6

[Alec Flett]

ok, this looks good to me - I'm no mac guy but sr=alecf

Comment 7

[Samir Gehani]

r=ssu

Comment 8

[Viswanath Ramachandran]

adding syd to comment on the beta stopperness of this bug. 

Comment 9

[Samir Gehani]

CC'ing folks with concerns about whether this is a stopper and if it is
necessary to take a code change at this point.  Also cc'ing folks the decision
to not take this will impact.

We have three options to ship something that point at the correct URL (which no
doubt everyone agrees we have to do):

1> Revert to early-2001 release madness of two separate installer builds
pointing to internal and external URLs.  IMHO, a step in the wrong direction
which requires work anyways.  There is associated risk and no build
infratrsucture for this anymore.  Needless to say, I can't make an estimate for
how long this would take for the release engineering folks and whether they are
in a position to commit the cycles.

2> Make the Site Selector section work (but Site Selector is supposed to be
disabled).  This also has risk and requires a code change.  So we'll get risk
with a fix that we won't want eventually down the line anyways.

3> Make global URL failovers work.  There is risk associated with this since
there is a code change.  However, this is the fix we want down the line anyways.

Since the risk assessment is similar for all three options, I hold that we
should take the fix for option 3, and do the right thing this time around once
and for all.

Comment 10

[J.J. Enser]

Samir is probably right that the Right Thing to do is probably to go with <3>, 
but just fyi, option <1> is no work at all for Release since the code is still in 
the automation (at least for mac) and only requires to flip one switch in order 
to have 2 stubs delivered with each verification build cycle, one being in a 
"external-ftp" sub-directory of each build repository.

Comment 11

[Jon Granrose]

I was under the impression that we were going with option #1  since that's what
we've done for previous releases and as jj states the build automation is set up
to do that.  I would bring this up with pdt, but expect that we will go with
option 1 for 0.9.1 and the global urls for 0.9.2.

and just a reminder that this is a mozilla bug in bugzilla and open to the world.

Comment 12

[Syd Logan]

I'd be inclined to go with option 3 since 0.9.1 is a better time to deal with 
issues than 0.9.2.

Comment 13

[Sean Su]

Given the time at hand, I think we should do 1) for the BRANCH because we've 
done it before and know that it works.

We can also do 3) for the TRUNK right away and set the urls appropriately so 
that they will fail over to the internal servers.  This should give us the 
failover testing that we want.  If this happens to slow QA down for the TRUNK 
testing, we can turn off the failover feature after adequately testing it.

Comment 14

[Samir Gehani]

FYI, we have not done option 1 before because pause/resume landed and the URL
logic changed.  Anyways, we have resolved that this will only go on the trunk.

Comment 15

[Christopher Blizzard (:blizzard)]

a=blizzard for the trunk on behalf of drivers

Comment 16

[Samir Gehani]

Fixed on trunk.  

Comment 17

[J.J. Enser]

whoever verifies this (Grace, Ktrina), should be aware that the fix didn't make 
to the 4am build (2001-06-22-04-trunk), but should be in 2001-06-22-08-trunk

Comment 18

[K'Trina Medina]

Verified