997389 - Should choose 32-bit builds by default on Windows

Status: RESOLVED WONTFIX

(Testing :: mozregression, defect)

Description

[William Lachance (:wlach)]

Right now you have to explicitly specify --bits=32 on Windows, which is a bit of a gotcha. We should just default to 32 on that particular platform.

Comment 1

[Dhruv Mullick]

Hello, I want to work on this bug. Please assign me to it.

Comment 2

[Dhruv Mullick]

Attachment: 32bitdefault.patch

A message is now displayed when a 64bit Win machine is detected and a 32bit installer is chosen by default.

Comment 3

[Aaron Train [:aaronmt]]

Comment on attachment 8409359 [details] [diff] [review]
32bitdefault.patch

Review of attachment 8409359 [details] [diff] [review]:
-----------------------------------------------------------------

Thanks for the patch Dhruv. I'm going to defer this over to William. I don't think this is what this patch is requesting in this bug.

Also we'll probably want this in the form of a pull-request over the repository: https://github.com/mozilla/mozregression

Comment 4

[William Lachance (:wlach)]

Comment on attachment 8409359 [details] [diff] [review]
32bitdefault.patch

Hi, thanks for the patch, but this isn't the quite the right way of fixing the problem. This patch is in a lower-level API (which only applies to inbound builds, not nightlies which we also care about), whereas we just want to change the default in the interface to running mozregression. 

For that, we currently we set the default number of bits here:

https://github.com/mozilla/mozregression/blob/master/mozregression/regression.py#L234

The trick is to change that value to a dynamically determined variable which we'll define a bit earlier: let's call that "defaultBits". The value of "defaultBits" should be equal to 32 (always) on Windows, on other platforms mozinfo.bits as it is now.

I'm happy to accept either pull requests or patches for mozregression. The former is slightly more convenient though.

Comment 5

[William Lachance (:wlach)]

Hi Dhruv, do you need any more information than that described in comment 4 to fix this bug? Let me know, happy to help in any way that I can.

Comment 6

[Rishabh Roy]

hi can i work on this bug ?
i have figured out a way how i can determine the bits of windows if it is 64 bit or 32 bit

Comment 7

[William Lachance (:wlach)]

(In reply to Rishabh Roy from comment #6)
> hi can i work on this bug ?
> i have figured out a way how i can determine the bits of windows if it is 64
> bit or 32 bit

Yup, it seems like Dhruv is not available right now so I've reassigned the bug to you. Note that the trick isn't to determine whether windows is running 64-bit or not (we can already do that through mozinfo) but to unconditionally make mozregression download a 32-bit build regardless of the nature of the platform on Windows.

Comment 8

[Rishabh Roy]

hi william on which irc channel can i find you ?

Comment 9

[Sam Garrett]

(In reply to Rishabh Roy from comment #8)
> hi william on which irc channel can i find you ?

The Mozilla IRC details are here: https://wiki.mozilla.org/IRC

wlach and the rest of the ateam people will be on the #ateam channel.

Comment 10

[William Lachance (:wlach)]

Hey Rishabh, did you still want to work on it? Do you need any additional info?

Comment 11

[Kapil Singh [:kapy]]

Hello, is this bug for Windows platform of linux platform.
I would like to work on this bug.

Thanks
Kapil

Comment 12

[William Lachance (:wlach)]

(In reply to Kapil Singh [:kapy] from comment #11)
> Hello, is this bug for Windows platform of linux platform.
> I would like to work on this bug.
> 
> Thanks
> Kapil

Hey Kapil! Since it actually looks like we're going to be shipping Win64 Firefox soon, we should change mozregression to handle it instead of changing the behaviour as outlined in this bug. I filed bug 1085433 to handle this, please have a look and let me know if you're interested over there.