Build Breakage

Build Breakage

[Denis Kenzior]

[-- Attachment #1: Type: text/plain, Size: 703 bytes --]

The latest gisi changes break the build:

pep.c: In function ‘g_isi_pep_create’:
pep.c:80: error: ‘SOCK_NONBLOCK’ undeclared (first use in this function)
pep.c:80: error: (Each undeclared identifier is reported only once
pep.c:80: error: for each function it appears in.)
pep.c:80: error: ‘SOCK_CLOEXEC’ undeclared (first use in this function)
make[2]: *** [pep.lo] Error 1
make[1]: *** [all-recursive] Error 1

SOCK_NONBLOCK and SOCK_CLOEXEC are only available on kernel >= 2.6.28 & glibc 
>= 2.9.  I know this is hardly bleeding edge, but oFono should continue 
working with older systems.  Perhaps a --disable-isi configure option is 
_really_ in order?

Regards,
-Denis

Re: Build Breakage

[=?unknown-8bit?q?R=C3=A9mi?= Denis-Courmont]

[-- Attachment #1: Type: text/plain, Size: 699 bytes --]


On Wed, 19 Aug 2009 19:59:37 -0500, Denis Kenzior <denkenz@gmail.com>
wrote:
> The latest gisi changes break the build:
> 
> pep.c: In function ‘g_isi_pep_create’:
> pep.c:80: error: ‘SOCK_NONBLOCK’ undeclared (first use in this
> function)
> pep.c:80: error: (Each undeclared identifier is reported only once
> pep.c:80: error: for each function it appears in.)
> pep.c:80: error: ‘SOCK_CLOEXEC’ undeclared (first use in this
> function)
> make[2]: *** [pep.lo] Error 1
> make[1]: *** [all-recursive] Error 1
> 
> SOCK_NONBLOCK and SOCK_CLOEXEC are only available on kernel >= 2.6.28 &
> glibc 

kernel <= 2.6.27 do not support ISI...

-- 
Rémi Denis-Courmont

Re: Build Breakage

[=?unknown-8bit?q?R=C3=A9mi?= Denis-Courmont]

[-- Attachment #1: Type: text/plain, Size: 693 bytes --]

On Wed, 19 Aug 2009 19:59:37 -0500, Denis Kenzior <denkenz@gmail.com>
wrote:
> The latest gisi changes break the build:
> 
> pep.c: In function ‘g_isi_pep_create’:
> pep.c:80: error: ‘SOCK_NONBLOCK’ undeclared (first use in this
> function)
> pep.c:80: error: (Each undeclared identifier is reported only once
> pep.c:80: error: for each function it appears in.)
> pep.c:80: error: ‘SOCK_CLOEXEC’ undeclared (first use in this
> function)
> make[2]: *** [pep.lo] Error 1
> make[1]: *** [all-recursive] Error 1
> 
> SOCK_NONBLOCK and SOCK_CLOEXEC are only available on kernel >= 2.6.28 &
> glibc >= 2.9.

Attached patch should fix it.

-- 
Rémi Denis-Courmont

[-- Attachment #2: 0001-gisi-PEP-use-more-portable-non-blocking-and-close.patch --]
[-- Type: text/plain, Size: 958 bytes --]

From 3ab2402ebc41204a53947aead4739ff92c1d46ff Mon Sep 17 00:00:00 2001
From: =?utf-8?q?R=C3=A9mi=20Denis-Courmont?= <remi.denis-courmont@nokia.com>
Date: Thu, 20 Aug 2009 10:52:57 +0300
Subject: [PATCH] gisi: PEP: use more portable non-blocking and close-on-exec

---
 gisi/pep.c |    5 ++++-
 1 files changed, 4 insertions(+), 1 deletions(-)

diff --git a/gisi/pep.c b/gisi/pep.c
index 89c6a80..abfe106 100644
--- a/gisi/pep.c
+++ b/gisi/pep.c
@@ -77,10 +77,13 @@ GIsiPEP *g_isi_pep_create(GIsiModem *modem)
 	unsigned ifi = g_isi_modem_index(modem);
 	char buf[IF_NAMESIZE];
 
-	fd = socket(PF_PHONET, SOCK_SEQPACKET|SOCK_NONBLOCK|SOCK_CLOEXEC, 0);
+	fd = socket(PF_PHONET, SOCK_SEQPACKET, 0);
 	if (fd == -1)
 		return NULL;
 
+	fcntl(fd, F_SETFD, FD_CLOEXEC);
+	fcntl(fd, F_SETFL, O_NONBLOCK|fcntl(fd, F_GETFL));
+
 	if (if_indextoname(ifi, buf) == NULL ||
 	    setsockopt(fd, SOL_SOCKET, SO_BINDTODEVICE, buf, IF_NAMESIZE))
 		goto error;
-- 
1.6.0.4

Re: Build Breakage

[Aki Niemi]

[-- Attachment #1: Type: text/plain, Size: 1080 bytes --]

2009/8/20 Denis Kenzior <denkenz@gmail.com>:
> SOCK_NONBLOCK and SOCK_CLOEXEC are only available on kernel >= 2.6.28 & glibc
>>= 2.9.  I know this is hardly bleeding edge, but oFono should continue
> working with older systems.  Perhaps a --disable-isi configure option is
> _really_ in order?

I think I've seen enough of these issues. I agree this is not nice, so
I'll push a patch for a --disable-isi configure switch, along with
Rémi's patch that fixes the actual build issue.

But it seems the build is broken in git, but for a different reason:

atmodem/atmodem.c:40:35: error: ofono/message-waiting.h: No such file
or directory
atmodem/atmodem.c:42:23: error: ofono/sim.h: No such file or directory
atmodem/atmodem.c:43:23: error: ofono/sms.h: No such file or directory
atmodem/atmodem.c:46:29: error: ofono/voicecall.h: No such file or directory
atmodem/atmodem.c: In function ‘create_cb’:
atmodem/atmodem.c:381: warning: assignment makes pointer from integer
without a cast

Perhaps a --disable-at switch? I kid, I kid... ;-)

Cheers,
Aki

Re: Build Breakage

[Denis Kenzior]

[-- Attachment #1: Type: text/plain, Size: 974 bytes --]

Hi Aki,

> But it seems the build is broken in git, but for a different reason:
>
> atmodem/atmodem.c:40:35: error: ofono/message-waiting.h: No such file
> or directory
> atmodem/atmodem.c:42:23: error: ofono/sim.h: No such file or directory
> atmodem/atmodem.c:43:23: error: ofono/sms.h: No such file or directory
> atmodem/atmodem.c:46:29: error: ofono/voicecall.h: No such file or

I already complained about this to Marcel. Any time a header is added a make 
clean is required. There were lots of new headers added.

> directory atmodem/atmodem.c: In function ‘create_cb’:
> atmodem/atmodem.c:381: warning: assignment makes pointer from integer
> without a cast

This probably happens because the message-waiting header isn't found :)

>
> Perhaps a --disable-at switch? I kid, I kid... ;-)
>

I've actually already said this should be the case :)  If one is building for 
a Nokia device, atmodem plugin is wasted space.

Regards,
-Denis