882756 - [SMS][MMS] We don't find the correct contact if an incoming SMS has a local/national number a=npotb

Status: RESOLVED FIXED

(Firefox OS Graveyard :: Gaia::SMS, defect)

Description

[Hubert Figuiere [:hub]]

When I receive SMS from my partner, the SMS appears to be coming from "5145555555". When she calls, the call display show "15145555555" and then match the name in the address book. 1 is the N-A country code.

On Android, it also matches the name in the address book, even though the call log show a different entry because it sorts by phone numbers.

Also, in the case of N-A numbers, Android will format them 1-XXX-XXX-XXXX

Comment 1

[Rick Waldron [:rwaldron]]

Can you provide a numbered set of steps to reproduce? Thanks!

Comment 2

[Hubert Figuiere [:hub]]

1. Get two SIM card for a North American carrier, preferably in to different call areas (I'm sure it is reproducable without, but that's not my setup), preferably in Koodo and this *might* be important as I'm not sure what configures how the phone number is send on a SMS.

1a Get a Geeksphone Keon GP-002(fx) and something else like an iPhone 4S (iOS 5 in that setup. I'm sure and Android phone would do fine). Make sure the service include "call display" (usually a $7 / month option). If you really want to be accurate, get a 604 area code number for the Keon and a 514 area code for the iPhone. I'm sure it doesn't matter.

2. Enter both phone number into each address book. This will allow to have matching on name in the address book. These should be entered with the 1 prefix. This is needed if one want to dial between different area code (so that they charge you more)

3. From the iPhone, send a SMS to the Geeksphone.

4. From the Geeksphone view the SMS.

Expected: it at least match the SMS sender to the entry in the address book.

Observed: it doesn't match

5. From the iPhone, call the Geeksphone. Notice that the call display will show the name matching from the address book.

6. Now put the sim card from the Geeksphone into a regular Android phone (running ICS Samsung in my case).

Repeat 2, 3, 4 replacing "Geeksphone" with "Android"

Observed: the Android phone will match the entry in the address book.
The call log will show the sender phone number without the 1 prefix.

Comment 3

[Rick Waldron [:rwaldron]]

cc julienw

Julien, what do you think about this? This seems like it's outside of Messages app.

Comment 4

[Julien Wajsberg [:julienw]]

I don't get why we don't get the normalized number for Sms, but we still should be able to find it.

Gregor, any thought on this ?

Comment 5

[Rick Waldron [:rwaldron]]

Here's the rundown:

1. Numbers saved in contacts are saved as-is and is not normalized.
2. New messages to recipients by selecting a known contact, by either the contact search suggestions or by contact select activity, will result in _exactly_ that number (not normalized) being used in the participants list.
3. If a message thread originates from a received message (from a contact or not) the number in the participants list is normalized.

Comment 6

[Gregor Wagner [:gwagner]]

The contacts API should support this use-case but the SMS app has a separate DB for matching SMS conversations. I think vicamo is the right guy to ask here.
So just to make sure. This is b2g18 and we are talking about the SMS app.
Does the call identification work?

Comment 7

[Rick Waldron [:rwaldron]]

This will be trivial to resolve once bug 881248 has landed

Comment 8

[Vicamo Yang [:vicamo][:vyang]]

(In reply to Julien Wajsberg [:julienw] from comment #4)
> I don't get why we don't get the normalized number for Sms, but we still
> should be able to find it.

We can't control the TOA (Type-Of-Address) of an incoming SMS message, it's assigned by SMS service center and could be national/international/numeric/... etc. depending on the policy deployed.  For example, carriers in Taiwan always use international number format even your friend and you are both clients of the same carrier and he/she is sitting next to you while texting you a short message.

So obviously, to have a "normalized" SMS number displayed/stored in db brings some problems and that was actually what we had done but corrected in bug 833291.  Normalized number is not a right solution for phone number matching.  Instead, the method introduced in bug 833291 is a near approximation.

(In reply to Gregor Wagner [:gwagner] from comment #6)
> The contacts API should support this use-case but the SMS app has a separate
> DB for matching SMS conversations.

Yes, SMS db has a participants table to match national/internaltional number.  The thing we really need is to have similar, shared or in separate, function in Contacts db.

Comment 9

[Gregor Wagner [:gwagner]]

> 
> (In reply to Gregor Wagner [:gwagner] from comment #6)
> > The contacts API should support this use-case but the SMS app has a separate
> > DB for matching SMS conversations.
> 
> Yes, SMS db has a participants table to match national/internaltional
> number.  The thing we really need is to have similar, shared or in separate,
> function in Contacts db.

We have now a 'match' lookup mechanism (bug 862250) that should solve this problem. 
Lets chat about it next week in Taipei.

Comment 10

[Rick Waldron [:rwaldron]]

Hubert, can you check this issue with gaia/master? Thanks!

Comment 11

[Hubert Figuiere [:hub]]

No I can't because I can't build for the Keon and that's the only phone that works as a phone.

I have master on unagi but unagi doesn't work with my carrier.

Comment 12

[Rick Waldron [:rwaldron]]

Are builds broken?

Comment 13

[Rick Waldron [:rwaldron]]

copied from bug 883663

--------------------------
Can you at least provide a verbatim example of how the phone number appears in the contacts app? Does it look like these?

For example, the same US number can be written differently:

[phone]
9995551234
[phone]
1 [phone]
[phone]

Comment 14

[Gregor Wagner [:gwagner]]

You don't have to build it yourself. I think you can also install a nightly from http://downloads.geeksphone.com/

Comment 15

[Hubert Figuiere [:hub]]

(In reply to Rick Waldron from comment #12)
> Are builds broken?

I have no more disk space to add another build (for Keon).

ALSO since it is my daily phone, I don't feel confident installing a build from master. I have had bad experience when using my Unagi in European work weeks....

Comment 16

[Hubert Figuiere [:hub]]

(In reply to Gregor Wagner [:gwagner] from comment #14)
> You don't have to build it yourself. I think you can also install a nightly
> from http://downloads.geeksphone.com/

This will wipe all my settings. Not cool.

Comment 17

[Hubert Figuiere [:hub]]

(In reply to Rick Waldron from comment #13)
> copied from bug 883663
> 
> --------------------------
> Can you at least provide a verbatim example of how the phone number appears
> in the contacts app? Does it look like these?
> 
> For example, the same US number can be written differently:
> 
> [phone]
> 9995551234
> [phone]
> 1 [phone]
> [phone]

On which version? Comment 0 tell how they are currently displayed in 1.0.1 for each case. The address book has it with [phone]

Comment 18

[Rick Waldron [:rwaldron]]

I recently landed code that makes smarter contact number comparisons based on inferences derived from the ITU-T E.123 specification. I'm confident those changes have addressed the issue reported in this bug.

Comment 19

[Rick Waldron [:rwaldron]]

Attachment: Github Pull Request https://github.com/mozilla-b2g/gaia/pull/10472

Comment 21

[Rick Waldron [:rwaldron]]

Requesting leo+

This patch introduces 203 tests of our ITU-T E.123 "inspired" phone number matching tests, for locales:

- United States
- Germany
- France
- Spain
- Brazil
- Colombia
- Chili

This will help to catch future regressions in phone number matching


Risk: None

Comment 22

[Julien Wajsberg [:julienw]]

The patch is test only, we don't need leo+ to land it on v1-train. Just use "a=npotb" to uplift when it's reviewed and landed on master.

Comment 23

[Rick Waldron [:rwaldron]]

(In reply to Julien Wajsberg [:julienw] from comment #22)
> The patch is test only, we don't need leo+ to land it on v1-train. Just use
> "a=npotb" to uplift when it's reviewed and landed on master.

Thanks. Is there a list or guide for these acronyms?

Comment 24

[Julien Wajsberg [:julienw]]

I don't exactly know, but you can use a=tests too ;) to me it's just a matter of whoever is looking at the git log can get a grasp of what's going on.

Ideally everything landing on the branch should have a a=; either a=<nickname> if it got an approval, a=leo+ if it's leo+, etc. But we're not really following this so strictly in gaia yet (I think they do in Gecko).

Comment 25

[Rick Waldron [:rwaldron]]

Landed https://github.com/mozilla-b2g/gaia/commit/469d2ed4d1ce7b39d6b6ecd37e5966de5fffafce

Comment 26

[Rick Waldron [:rwaldron]]

v1-train https://github.com/mozilla-b2g/gaia/commit/f26c625afa5e1e53cbe900f09b93108d88503a1d

Comment 27

[John Ford [:jhford] CET/CEST Berlin Time]

v1.1.0hd: f26c625afa5e1e53cbe900f09b93108d88503a1d