--=-=-=


Love <lha@stacken.kth.se> writes:

> "Shin'ichiro TAYA" <taya@ba2.so-net.ne.jp> writes:
>
>> From: Love <lha@stacken.kth.se>
>> Subject: Re: CVS commit: pkgsrc/www/mozilla
>> Date: Mon, 21 Jun 2004 18:13:56 +0200
>>
>>> Thanks, how about the patch (patch-ab) below so mozilla at least
>>> configure correctly for res_ninit ? (dunno what the rest of the code does
>>> the right thing)
>>
>> This will detects res_ninit() correctory and define HAVE_RES_NINIT,
>> but I couldn't find anythig that utilize HAVE_RES_NINIT.
>> 'fgrep -R HAVE_RES_NINIT . ' only reports configre.in and configure.
>> How do you think about this?
>
> id-utils tells me
> : lha@nutcracker ; gid HAVE_RES_NINIT res_ninit 
> netwerk/dns/src/nsHostResolver.cpp:37:#if defined(HAVE_RES_NINIT)
> netwerk/dns/src/nsHostResolver.cpp:134:        LOG(("calling res_ninit\n"));
> netwerk/dns/src/nsHostResolver.cpp:137:        return (res_ninit(&_res) == 0);
>
> So, they are used, dunno why fgrep doesn't pick it out.

And they are used wrong, mozilla assumes that it can reach the TLS'ed
res_state as _res and reset whenever it fell to. If it tries do do this on
netbsd, it will fall over in a abort().

So, its not a good idea to apply the patch, rather it should me made sure
that code-path are not used, by making sure RES_RETRY_ON_FAILURE is not
defined in mozilla/netwerk/dns/src/nsHostResolver.cpp.

Love


--=-=-=
Content-Type: application/pgp-signature

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (NetBSD)

iQEVAwUAQNfwonW+NPVfDpmCAQL3+ggAiEQvj7bc11Ps1BEKG5abF9DvCRps7Wiq
OELVnxhNHTNiZuq3Vc1E2mZiI/wxmwAsAoRx0fO9EhAKIDujGJCDQRAzIBNvN4JR
dipjJ/MidI/WbrFfVF2ciAqfAwtNxTlgOfhr8h3MDBIwfPhEbgKV5JoytvK85BHh
dui++SdupCRy0WxwD4+gDE0NDs/pWSMhCPL6zaEDXRSgHU4TBpUbQQYJ87vUQczK
fprlol3BHrT0ueezqGDy/90t/ldFUBPxlyD7V/dN9vrq72Lwlt46bdOrqATl+14w
7d9YSzjKYJrEg1wugGYOmKicfOHlpWjr004QDqYZmCM6abc8nR31eg==
=vZGW
-----END PGP SIGNATURE-----
--=-=-=--