903830 - dom/network/tests/unit/test_tcpserversocket.js uses a hardcoded port

Status: RESOLVED INVALID

(Core :: DOM: Core & HTML, defect, P5)

Description

[Mihnea Dobrescu-Balaur (:mihneadb)]

This causes (intermittent) trouble when running the xpcshell tests in parallel.

[self reminder] Must remove the run-sequentially annotation introduced in bug 898658 when this gets fixed.

Comment 1

[Mihnea Dobrescu-Balaur (:mihneadb)]

Removing this as a dep because I'm having a bit of trouble with the interface. Will just mark the test to run-sequentially for now in bug 887054.

I see in the .idl file that the port parameter is an *unsigned* and that's not good since I have to pass in -1 to get a dynamic port (as the docstring actually suggests).

Also, I tried accessing the .port attribute (it's in the .idl as well) and the test just hangs.

Could use some help on this! :)

Thanks

Comment 2

[Donovan Preston [:fzzzy]]

Isn't port 0 the port to use to get a dynamic port, not -1? Can you point me to the docstring you are talking about?

Using port 0 and then accessing the .port attribute should be the correct way to get this to run in parallel. I will put an entry on my todo list to look into why accessing the .port attribute hangs.

Comment 3

[Mihnea Dobrescu-Balaur (:mihneadb)]

(In reply to Donovan Preston from comment #2)
> Isn't port 0 the port to use to get a dynamic port, not -1? Can you point me
> to the docstring you are talking about?
> 
> Using port 0 and then accessing the .port attribute should be the correct
> way to get this to run in parallel. I will put an entry on my todo list to
> look into why accessing the .port attribute hangs.

0 is in BSD/kernel land, yes. However, [1] says -1.

[1] http://dxr.mozilla.org/mozilla-central/source/dom/network/interfaces/nsIDOMTCPSocket.idl#l60

Comment 4

[Donovan Preston [:fzzzy]]

(In reply to Mihnea Dobrescu-Balaur (:mihneadb) from comment #3)
> 0 is in BSD/kernel land, yes. However, [1] says -1.
> 
> [1]
> http://dxr.mozilla.org/mozilla-central/source/dom/network/interfaces/
> nsIDOMTCPSocket.idl#l60

Hmm, you're right. I'm not sure I like that, I think I would prefer it to be 0. But that's not really important, the important thing is to figure out why passing -1 and then accessing .port hangs. I'll try to take a look tomorrow.

Comment 5

[Mihnea Dobrescu-Balaur (:mihneadb)]

(In reply to Donovan Preston from comment #4)
> (In reply to Mihnea Dobrescu-Balaur (:mihneadb) from comment #3)
> > 0 is in BSD/kernel land, yes. However, [1] says -1.
> > 
> > [1]
> > http://dxr.mozilla.org/mozilla-central/source/dom/network/interfaces/
> > nsIDOMTCPSocket.idl#l60
> 
> Hmm, you're right. I'm not sure I like that, I think I would prefer it to be
> 0. But that's not really important, the important thing is to figure out why
> passing -1 and then accessing .port hangs. I'll try to take a look tomorrow.

That's great, thanks!

Comment 6

[Donovan Preston [:fzzzy]]

Ok, I think I have a plan for fixing this. Testing now to see whether it will work.

Comment 7

[Donovan Preston [:fzzzy]]

Ok, so first of all, the docs are wrong. You actually pass 0 to have the OS choose a free port, and it actually works right now.

However, there is another bug, which is that it is not possible to know the port that was chosen when in a child process.

I believe the port can't be discovered by the child process synchronously, either, so there will have to be a new "onlistening" event or something like that which will indicate the port is now known to the child process.

Comment 8

[Mihnea Dobrescu-Balaur (:mihneadb)]

If the child process shares code with the parent you can inject it from there.

If it is too complicated you can just keep it as is (run-sequentially) and all will be fine.

Comment 9

[Donovan Preston [:fzzzy]]

Attachment: fix-localPort.diff

Here's a bugfix for a typo which was preventing access to the localPort attribute.

Comment 10

[Firefox Bug Husbandry Bot]

https://bugzilla.mozilla.org/show_bug.cgi?id=1472046

Move all DOM bugs that haven’t been updated in more than 3 years and has no one currently assigned to P5.

If you have questions, please contact :mdaly.

Comment 11

[Hsin-Yi Tsai (she/her) [:hsinyi]]

The referenced files in the previous comments such as nsIDOMTCPSocket.idl or TCPServerSocket.js are no longer in the codebase.