Good news, anyway.
Attached alsa-info.txt with model=b1900.

And I followed to use only data for automute while set_gpio inside 
init_verb; which works. Below is the code (attached diff):

/* toggle speaker-output according to the hp-jack state */
static void alc260_b1900_automute(struct hda_codec *codec)
{
unsigned int present;

present = snd_hda_jack_detect(codec, 0x0f);
if (present) {
snd_hda_codec_write_cache(codec, 0x01, 0,
AC_VERB_SET_GPIO_DATA,
1);
} else {
snd_hda_codec_write_cache(codec, 0x01, 0,
AC_VERB_SET_GPIO_DATA,
0);
}
}

static const struct hda_verb alc260_b1900_verbs[] = {
{0x0f, AC_VERB_SET_PIN_WIDGET_CONTROL, PIN_HP},
{0x0b, AC_VERB_SET_CONNECT_SEL, 0x00},
{0x0d, AC_VERB_SET_CONNECT_SEL, 0x00},
{0x0f, AC_VERB_SET_EAPD_BTLENABLE, 0x02},
{0x1a, AC_VERB_SET_COEF_INDEX, 0x07},
{0x1a, AC_VERB_SET_PROC_COEF, 0x3040},
{0x01, AC_VERB_SET_GPIO_MASK, 0x01},
{0x01, AC_VERB_SET_GPIO_DIRECTION, 0x01},
{0x0f, AC_VERB_SET_UNSOLICITED_ENABLE, AC_USRSP_EN | ALC880_HP_EVENT},
{}
};


On 2012年02月16日 17:42, Takashi Iwai wrote:
> At Thu, 16 Feb 2012 10:41:32 +0100,
> Takashi Iwai wrote:
>> At Thu, 16 Feb 2012 17:38:33 +0800,
>> joey.jiaojg wrote:
>>> YES, you are right on the assumption. I tried again many times of
>>> SET_GPIO_DATA between 0/1, speaker/headphone switches.
>>> So is possible to merge into next ALSA release?
>> Not ready for "merge".  As mentioned, your patch can't be applied to
>> the latest tree.
>>
>> I'll work on the better implementation of a quirk for the auto-parser
>> later.
> Oh, for that, could you give alsa-info.sh output?
> Attach the output file (don't paste) obtained with --no-upload
> option.
>
>
> thanks,
>
> Takashi