859779 - B2G Emulator-x86: build failure due to __NR_msgctl, __NR_msgsnd, __NR_msgrcv undeclared when compiling WebRTC

Status: RESOLVED FIXED

(Core :: WebRTC: Signaling, defect)

Description

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

gecko/media/webrtc/signaling/src/sipcc/cpr/linux/cpr_linux_ipc.c: In function 'msgsnd':
gecko/media/webrtc/signaling/src/sipcc/cpr/linux/cpr_linux_ipc.c:58: error: '__NR_msgsnd' undeclared (first use in this function)
gecko/media/webrtc/signaling/src/sipcc/cpr/linux/cpr_linux_ipc.c:58: error: (Each undeclared identifier is reported only once
gecko/media/webrtc/signaling/src/sipcc/cpr/linux/cpr_linux_ipc.c:58: error: for each function it appears in.)
gecko/media/webrtc/signaling/src/sipcc/cpr/linux/cpr_linux_ipc.c: In function 'msgrcv':
gecko/media/webrtc/signaling/src/sipcc/cpr/linux/cpr_linux_ipc.c:63: error: '__NR_msgrcv' undeclared (first use in this function)
gecko/media/webrtc/signaling/src/sipcc/cpr/linux/cpr_linux_ipc.c: In function 'msgctl':
gecko/media/webrtc/signaling/src/sipcc/cpr/linux/cpr_linux_ipc.c:68: error: '__NR_msgctl' undeclared (first use in this function)
gecko/media/webrtc/signaling/src/sipcc/cpr/linux/cpr_linux_ipc.c: In function 'msgget':
gecko/media/webrtc/signaling/src/sipcc/cpr/linux/cpr_linux_ipc.c:73: error: '__NR_msgget' undeclared (first use in this function)
gecko/media/webrtc/signaling/src/sipcc/cpr/linux/cpr_linux_ipc.c: In function 'cprCreateMessageQueue':
gecko/media/webrtc/signaling/src/sipcc/cpr/linux/cpr_linux_ipc.c:273: warning: implicit declaration of function 'ftok'

Comment 1

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

Attachment: patch

There is no __NR_msg{get,snd,rcv,ctl} syscalls in i386. Instead, __NR_ipc handles all these calls with an additional ID.

See:
http://www.eglibc.org/cgi-bin/viewvc.cgi/trunk/libc/sysdeps/unix/sysv/linux/ipc_priv.h?view=markup
http://www.eglibc.org/cgi-bin/viewvc.cgi/trunk/libc/sysdeps/unix/sysv/linux/i386/msgctl.c?view=markup
http://www.eglibc.org/cgi-bin/viewvc.cgi/trunk/libc/sysdeps/unix/sysv/linux/msgget.c?view=markup
http://www.eglibc.org/cgi-bin/viewvc.cgi/trunk/libc/sysdeps/unix/sysv/linux/msgrcv.c?view=markup
http://www.eglibc.org/cgi-bin/viewvc.cgi/trunk/libc/sysdeps/unix/sysv/linux/msgsnd.c?view=markup

Comment 2

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

https://tbpl.mozilla.org/?tree=Try&rev=0a53cc2dcb82

Comment 3

[Randell Jesup [:jesup] (needinfo me)]

Comment on attachment 735164 [details] [diff] [review]
patch

Review of attachment 735164 [details] [diff] [review]:
-----------------------------------------------------------------

Other than I don't know what the 5, 11 mean for __NR_ipc (which I don't see anywhere else in our tree; I'm guessing they're IPC channel identifiers), this seems reasonable.  I advise someone who knows where __NR_ipc comes from should review that the calls are correct.  From the webrtc side, this seems ok.

Comment 4

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

(In reply to Randell Jesup [:jesup] from comment #3)
> Comment on attachment 735164 [details] [diff] [review]
> Review of attachment 735164 [details] [diff] [review]:
> -----------------------------------------------------------------
> Other than I don't know what the 5, 11 mean for __NR_ipc (which I don't see
> anywhere else in our tree; I'm guessing they're IPC channel identifiers),
> this seems reasonable.  I advise someone who knows where __NR_ipc comes from
> should review that the calls are correct.  From the webrtc side, this seems
> ok.

5 means there are 5 more parameters; 11, 12, 13, 14 are opcodes, eglibc defines them in ipc_priv.h. I should have done the same to make it more clear:

  #define IPCOP_msgsnd	11
  #define IPCOP_msgrcv	12
  #define IPCOP_msgget	13
  #define IPCOP_msgctl	14

Comment 5

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

Attachment: v2

1) include Android system headers for constants and structs.
2) Removed magic number '5' for it's actually a redundent macro variable in eglibc
3) From the definition of struct ipc_perm defined in bionic/libc/kernel/common/linux/ipc.h[1], Android is using old ABI. So mask cmd parameter in msgctl explicitly.

@dhylands, could you help review this patch?

[1]: https://github.com/mozilla-b2g/platform_bionic/blob/ics-mr1-release/libc/kernel/common/linux/ipc.h#L19

Comment 6

[Dave Hylands [:dhylands]]

Comment on attachment 736160 [details] [diff] [review]
v2

Review of attachment 736160 [details] [diff] [review]:
-----------------------------------------------------------------

Although I have to ask the bigger question. Why aren't we using the same IPC mechanisms used in say gecko for communicating between the child and parent processes?

Comment 7

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

(In reply to Dave Hylands [:dhylands] from comment #6)
> Although I have to ask the bigger question. Why aren't we using the same IPC
> mechanisms used in say gecko for communicating between the child and parent
> processes?

Have no idea here. Just fix compile errors for emulator-x86 build :(

Comment 8

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

Attachment: v3

Add r= only

Comment 9

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

https://hg.mozilla.org/integration/mozilla-inbound/rev/36cd103a646d

Comment 10

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/36cd103a646d