102118 - solaris 2.7, gcc 3.0.1 cannot convert `int*' to `socklen_t*' calling getsockname

Status: VERIFIED FIXED

(SeaMonkey :: Installer, defect)

Description

[Nils Lohner]

From Bugzilla Helper:
User-Agent: Mozilla/5.0 (X11; U; SunOS sun4u; en-US; rv:0.9.3) Gecko/20010821
BuildID:    0.9.4


make[2]: Entering directory `/sup/build/mozilla/mybuild/xpinstall/wizard/libxpnet'
make[3]: Entering directory
`/sup/build/mozilla/mybuild/xpinstall/wizard/libxpnet/src'
nsSocket.cpp
c++ -o nsSocket.o -c -DOSTYPE=\"SunOS5\" -DOSARCH=\"SunOS\" -DOJI  
-I../../../../dist/include -I../../../../dist/include
-I/sup/build/mozilla/mybuild/dist/include/nspr      -I/usr/openwin/include  
-fPIC  -I/usr/openwin/include -fno-rtti -fno-exceptions -pedantic -Wno-long-long
-pipe -fshort-wchar -pthreads -O  -DNDEBUG -DTRIMMED -I/usr/openwin/include
-DMOZILLA_CLIENT -include ../../../../config-defs.h -Wp,-MD,.deps/nsSocket.pp
../../../../../xpinstall/wizard/libxpnet/src/nsSocket.cpp
../../../../../xpinstall/wizard/libxpnet/src/nsSocket.cpp: In member function 
   `int nsSocket::SrvAccept()':
../../../../../xpinstall/wizard/libxpnet/src/nsSocket.cpp:247: cannot convert 
   `int*' to `socklen_t*' for argument `3' to `int accept(int, sockaddr*, 
   socklen_t*)'
../../../../../xpinstall/wizard/libxpnet/src/nsSocket.cpp: In member function 
   `int nsSocket::GetHostPortString(char**)':
../../../../../xpinstall/wizard/libxpnet/src/nsSocket.cpp:456: cannot convert 
   `int*' to `socklen_t*' for argument `3' to `int getsockname(int, sockaddr*, 
   socklen_t*)'
make[3]: *** [nsSocket.o] Error 1
make[3]: Leaving directory
`/sup/build/mozilla/mybuild/xpinstall/wizard/libxpnet/src'
make[2]: *** [install] Error 2
make[2]: Leaving directory `/sup/build/mozilla/mybuild/xpinstall/wizard/libxpnet'
make[1]: *** [install] Error 2
make[1]: Leaving directory `/sup/build/mozilla/mybuild/xpinstall'
make: *** [install] Error 2
4575.30u 410.88s 2:02:23.56 67.8%
/sup/build/mozilla/mybuild > 


... #mozilla irc.mozilla.org

<Pike> WHAT?
<Pike> #if defined(SOLARIS) || defined(_WINDOWS)
<Pike> #define socklen_t int
<Pike> #endif
<Pike> that must be OLD cruft solaris
<Pike> 'cause on 2.7, the function prototype is ok
<Pike>      int  accept(int  s,   struct   sockaddr   *addr,   socklen_t
<Pike>      *addrlen);
<Pike> I don't seem to have SOLARIS defined here
<Pike> CQ: I do see why this errors
<Pike> I just wonder why it's ok for me
<Pike> now I know
<Pike> hey
<Pike> how do I get gcc to tell it's standard defines once again?
<Pike> CQ: you should file a bug, IMHO
<Pike> CQ: CC [e-mail] and [e-mail]
<jag> CQ: maybe cc [e-mail] too
<Pike> solaris headers are evil
<CQ> pike- will do.  How do I continue my build though?  Just nuke those three
lines?
<Pike> typedef uint32_t        socklen_t;
<Pike> typedef void            *Psocklen_t;
<Pike> isn't this a crying game?
<Pike> CQ: I honestly don't know. But if you can do something to make it compile
for you, it should be ok
<CQ> pike- OK if I include your IRC comments in teh bug report?
<Pike> sure
<Pike> though I do know now why I have SOLARIS defined
<CQ> Pike- why?
<CQ> ..and which componendt should I file this against?
<Pike> CQ: it's in platform.h
<Pike> CQ: and I have no clue on what component. It's xpinstall, but it's
probably a netwerk issue, and should be resolved by some build config foo
<Pike> CQ: file it on xpinstall for now


Reproducible: Always
Steps to Reproduce:
1. ./configure --disable-tests --disable-debug --enable-optimize --enable-crypto
--disable-xprint --with-gnu-ld --with-gnu-nm --with-gnu-as --with-as=/sup/bin/as
--with-ld=/sup/bin/ld --with-nm=/sup/bin/nm --prefix=/sup
2. make

/sup/build/mozilla/mybuild > uname -a
SunOS lou 5.7 Generic_106541-15 sun4u sparc SUNW,Ultra-5_10
/sup/build/mozilla/mybuild > gcc --version
3.0.1

Comment 1

[timeless]

fwiw _WINDOWS shouldn't survive whatever changes are made.

Comment 2

[Daniel Veditz [:dveditz]]

This isn't the XPInstall engine, it's the wizard socket library.

I see a lot of folks CC'd -- anyone want to confirm the bug?

Comment 3

[hacker formerly known as [e-mail]]

*** Bug 104203 has been marked as a duplicate of this bug. ***

Comment 4

[Brice Lecomte]

From <sys/socket.h>
on solaris 8
#ifndef
_SOCKLEN_T
#define
_SOCKLEN_T

#if defined(_XPG4_2) && !defined(_XPG5) && !defined(_LP64)
typedef
size_t
	socklen_t;
#else
typedef
uint32_t
socklen_t;
#endif
/* defined(_XPG4_2) && !defined(_XPG5) && !defined(_LP64) */

#if defined(_XPG4_2)
typedef
socklen_t
*Psocklen_t;
#else
typedef
void
	*Psocklen_t;
#endif
/* defined(_XPG4_2) */

#endif
/* _SOCKLEN_T */

so which one to choose ? size_t* or uint_32_t* to cast in function call ?
int accept(int, sockaddr*, socklen_t*)
int getsockname(int, sockaddr*, socklen_t*)

of nsSocket.cpp: In member function `int nsSocket::SrvAccept()'

Comment 5

[jap3003+bugzilla]

I was getting this one back with gcc 3.0 (i.e. 3.0.1)  [oh, yeah.  Solaris 2.8
07/01].  Thinking it might be a bug in GCC (since it works for both Forte
Developer 6.2 and for gcc 2.95.3), I filed it with GCC.  They got back to me and
told me it's a Mozilla bug.  So that's where I was until today.

Both size_t and uint_32_t seem to work.  I haven't the foggiest what the _XPG4_2
and _XPG5 symbols are supposed to represent.  Ah.  There they are. 
http://ftp.planetmirror.com/pub/qt/mkspecs says they're Open Group 4v2 and 5 specs.

Hrm.  Is there any reason we're defining socklen_t to be an int?  It compiles
just fine without this, and I'd think the OS would know best what it understands
as a socklen_t.

Comment 6

[jap3003+bugzilla]

Erm, that should be uint32_t, not uint_32_t.  Ewps.

Comment 7

[David Bradley]

Uros suggested that bug 103499 may be a dupe of this. If not is it at least
tackling similar problems and might be of interest to people here.

Comment 8

[Jon Smirl]

*** Bug 103499 has been marked as a duplicate of this bug. ***

Comment 9

[Brice Lecomte]

with 0.9.5 on Solaris 8, with (size_t*) as cast of the 5th element ll 163,168
of nsSocket.cpp, it compiles with gcc301 binutils2.11.2 glib 1.2.10 gtk1.2 and
the required patches but 

*** Chrome Registration of package: Checking for contents.rdf at
jar:resource:/chrome/comm.jar!/content/necko/
###!!! ASSERTION: Cannot tell if this is a directory: 'NS_SUCCEEDED(rv)', file
nsLocalFileUnix.cpp, line 1436
###!!! Break: at file nsLocalFileUnix.cpp, line 1436

*** Chrome Registration of locale: Checking for contents.rdf at
jar:resource:/chrome/US.jar!/locale/US/messenger-region/
###!!! ASSERTION: never called: 'Error', file
../../../../dist/include/xpconnect/xpc_map_end.h, line 143
###!!! Break: at file ../../../../dist/include/xpconnect/xpc_map_end.h, line 143
Segmentation Fault - core dumped

Comment 10

[Roland Mainz]

Any stack trace (hint: try "/usr/proc/bin/pstack <core-file>" ...) ?

Comment 11

[Uros Bizjak]

Assertion in xpc_map_end.h, line 143 is caused by bug 71627. I have changed
xptcinvoke_asm_sparc_solaris_GCC.s to xptcinvoke_asm_sparc_solaris_GCC3.s as
suggested in bugreport, and I was able to run CVS mozilla on solaris 2.8,
compiled with gcc 3.0.2.

Comment 12

[hacker formerly known as [e-mail]]

*** Bug 110819 has been marked as a duplicate of this bug. ***

Comment 13

[Brice Lecomte]

* To Uros Bizjak : I cannot use cvs so I get source.tar.gz and TRY to compile...
I hope last file version is put in the tarball !
* to jap3003 (and others ;-):
I still needed to cast xpinstall/wizard/libxpnet/src/nsSocket.cpp as
163:getsockopt( mFd, SOL_SOCKET, SO_RCVBUF, (char*) &newTCPWin,(Psocklen_t)&len );

168:getsockopt( mFd, SOL_SOCKET, SO_RCVBUF, (char*) &newTCPWin,(Psocklen_t)&len );

247:mFd = accept(mListenFd, (struct sockaddr *) &cliaddr, (Psocklen_t)&clilen);
456:if ((getsockname(mListenFd, (struct sockaddr *) &servaddr,
(Psocklen_t)&salen)) < 0)

from mozilla-source.tar.gz 30th Nov
it runs with basic (no) configure options (any optimization crashe)

Comment 14

[Daniel Veditz [:dveditz]]

*** Bug 126556 has been marked as a duplicate of this bug. ***

Comment 15

[Jim Crumley]

*** Bug 128250 has been marked as a duplicate of this bug. ***

Comment 16

[Henrik Johansson]

Attachment: Same problem in 0.9.8 and 0.9.9 on Solaris 7 and 8

Comment 17

[hacker formerly known as [e-mail]]

Attachment: Check for SOLARIS && !_SOCKLEN_T

Sparc/solaris 2.7, at least, provides a define if it provides socklen_t so we
should probably use it.

Comment 18

[Daniel Veditz [:dveditz]]

nominating for mozilla1.0 -- not building on Solaris is pretty stupid. Handing
off to seawoood since he's got a handle on the problem and is the master of the
tbox machines.

r=dveditz for the patch

Comment 19

[Alec Flett]

Comment on attachment 77408 [details] [diff] [review]
Check for SOLARIS && !_SOCKLEN_T

sr=alecf

Comment 20

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

Comment on attachment 77408 [details] [diff] [review]
Check for SOLARIS && !_SOCKLEN_T

a=[e-mail] for drivers

Comment 21

[hacker formerly known as [e-mail]]

bbaetz checked this in yesterday.

Comment 22

[K'Trina Medina]

Verified code fix

Comment 23

[timeless]

*** Bug 136494 has been marked as a duplicate of this bug. ***

Comment 24

[Daniel Veditz [:dveditz]]

*** Bug 146547 has been marked as a duplicate of this bug. ***