headers
Yu Shao | 2 Dec 2003 08:15
Picon
Favicon

[openi18n-im:00535] a binary package splitting scheme

Hi All,

May I suggest a binary package splitting scheme for im-sdk, since last 
time Hideki brought this issue up, there were some changes occurred, how 
do you think of following structure:

iiimf-server     (htt,  htt_server)
iiimf-libs         (libiiimp, libiiimcf, libEIMIL, csconv)
iiimf-x             (iiimxcf, httx, htt_xbe, xiiimp.so)
iiimf-java
iiimf-emacs
iiimf-gtk          (iiimgcf )
iiimf-qt            (iiimqcf )
iiimf-docs
iiimf-le-devel  (headers, libs needed for developing LEs)
iiimf-le-*         (per language, like iiimf-le-zh_cn, iiimf-le-zh-tw, 
iiimf-le-ja etc)

Thanks,
Yu Shao
Permalink | Reply | 5 comments |
headers
Yu Shao | 2 Dec 2003 08:44
Picon
Favicon

[openi18n-im:00536] Re: a one line htt patch

When checking the email archive today, I was sure, this patch will fix 
the bug reported by Rajkumar S about htt looping. Email No.00333.

Regards,
Shao
Yu Shao wrote:

> Just a one line patch in case people get infinite loop when start htt 
> server. It happened to me several times, when iml_if is not set, and 
> LEBase destructor is called, the code will get Seg fault when try to 
> deallocate iml_if.
>
> Regards,
> Shao
>
> --- LE.cpp.orig 2003-11-13 09:22:15.000000000 +1000
> +++ LE.cpp 2003-11-13 09:22:46.000000000 +1000
>  <at>  <at>  -227,6 +227,7  <at>  <at> 
> filename = x_filename;
> need_thread_lock = false;
> error = false;
> + iml_if = NULL;
>
> if (!loadif()) error = true;
> }
>
>
Permalink | Reply | 1 comment |
headers
David Joo | 2 Dec 2003 08:44
Picon
Favicon

[openi18n-im:00537] Re: a binary package splitting scheme


On Tue, 2 Dec 2003, Yu Shao wrote:

> Hi All,
> 
> May I suggest a binary package splitting scheme for im-sdk, since last 
> time Hideki brought this issue up, there were some changes occurred, how 
> do you think of following structure:
> 
> iiimf-server     (htt,  htt_server)
> iiimf-libs         (libiiimp, libiiimcf, libEIMIL, csconv)
> iiimf-x             (iiimxcf, httx, htt_xbe, xiiimp.so)
> iiimf-java
> iiimf-emacs
> iiimf-gtk          (iiimgcf )
> iiimf-qt            (iiimqcf )
> iiimf-docs
> iiimf-le-devel  (headers, libs needed for developing LEs)
> iiimf-le-*         (per language, like iiimf-le-zh_cn, iiimf-le-zh-tw, 
> iiimf-le-ja etc)
> 
> Thanks,
> Yu Shao
> 

--

-- 
ÁÖ¼º¹Î (David SungMin Joo)                                djoo <at> redhat.com
Red Hat Asia-Pacific Pty. Ltd.
Ph   :   +61 7 3872 4833  Fax: +61 7 3257 4800
Legal:   http://apac.redhat.com/disclaimer
(Continue reading)

Permalink | Reply | 0 comments |
headers
Roger So | 2 Dec 2003 09:53

[openi18n-im:00538] Re: a binary package splitting scheme

On Tue, 2003-12-02 at 15:15, Yu Shao wrote:
> Hi All,
> 
> May I suggest a binary package splitting scheme for im-sdk, since last 
> time Hideki brought this issue up, there were some changes occurred, how 
> do you think of following structure:
[...]
> iiimf-le-devel  (headers, libs needed for developing LEs)

This is definitely needed.  At least the header files IM*.h and
iml/Sun*.h need to be packaged in some way.

> iiimf-le-*         (per language, like iiimf-le-zh_cn, iiimf-le-zh-tw, 
> iiimf-le-ja etc)

I'm not sure if per-language is a good idea.  What if there are multiple
LEs available for a language, and the user only wants one installed?

I think one package per LE is a better approach.

Roger
--

-- 
  Roger So                 Community Representative
  Sun Wah Linux Limited    Chinese Platform Developer
  Tel: +852 2250 0230      roger.so <at> sw-linux.com
  Fax: +852 2259 9112      http://www.sw-linux.com/
Permalink | Reply | 0 comments |
headers
Hidetoshi Tajima | 2 Dec 2003 11:21
Picon

[openi18n-im:00539] Re: a binary package splitting scheme

Hi,

I heard from Hideki that since the conversion was discussed in the
alias before, debian's packages has been already made accordingly. Not?

If so, then why don't we follow the convension and extend it if
necessary?

Actually, I've created spec.in file for im-sdk trunk according to the
convensions. With this, we would package binary RPMs as follows:

 - iiimf-htt-csconv
 - iiimf-htt-le-canna 
 - iiimf-htt-le-unit
 - iiimf-htt-le-newpy 
 - iiimf-htt-server 
 - iiimf-htt-xbe 
 - iiimgcf 
 - im-sdk-docs 
 - libiiimcf2-devel
 - libiiimcf2
 - libiiimp0-devel
 - libiiimp0

can I go ahead and commit this, or shall I post to the list for review
first?

Regards,
Toshi
(Continue reading)

Permalink | Reply | 2 comments |
headers
Zarick Lau | 2 Dec 2003 11:01
Favicon

[openi18n-im:00540] A problem in use iml_make_commit_inst in a aux_event

Hi,

I have noticed a problem on htt_server! 

During if_SendEvent () call to le, if the event is IM_EventAux, the
iml_session_t is a special session (I believe it is a special session
for XAux communication)

When I try to call iml function (e.g. commit / env_conv) the call will
have no effect at all.

I have also try to save the last used session (which should a normal
session/ a valid input context) and calling iml_* with that saved
session pointer, but 
again, nothing happens.

What is the proper way to using iml_* function during a XAux event call?

Best Regards,
Zarick Lau
Permalink | Reply | 13 comments |
headers
Yu Shao | 3 Dec 2003 08:10
Picon
Favicon

[openi18n-im:00541] Re: a binary package splitting scheme

Hidetoshi Tajima 写道:

>Hi,
>
>I heard from Hideki that since the conversion was discussed in the
>alias before, debian's packages has been already made accordingly. Not?
>  
>
I did check the original thread and what I want is bringing more
feedback in.

>If so, then why don't we follow the convension and extend it if
>necessary?
>
>Actually, I've created spec.in file for im-sdk trunk according to the
>convensions. With this, we would package binary RPMs as follows:
>
> - iiimf-htt-csconv
> - iiimf-htt-le-canna 
> - iiimf-htt-le-unit
> - iiimf-htt-le-newpy 
> - iiimf-htt-server 
> - iiimf-htt-xbe 
>
I am not opposing using "htt", but if htt is the only iiimf
implementation, iiimf-server would be more friendly to normal users.
Also few normal users would like to figure out what "xbe" is(I know it
is easy to distingush it as developer) and then install it, and also
actually htt-xbe will be invoked by httx, why don't simply put them
together and call them iiimf-x(or iiimf-htt-x).
(Continue reading)

Permalink | Reply | 1 comment |
headers
Motonobu Ichimura | 2 Dec 2003 15:22
Favicon

[openi18n-im:00542] Re: a one line htt patch

Hi,
sorry for replying too late.

I've already committed this patch in main trunk.

-------------------------------------------------------------------
2003-11-19  Motonobu Ichimura  <famao <at> momonga-linux.org>

        * LE.cpp (LEBase::LEBase): fix a missing initialization for
iml_if.
        patch from Yu Shao <yshao <at> redhat.com>
--------------------------------------------------------------------

On Tue, 2003-12-02 at 16:44, Yu Shao wrote:
> When checking the email archive today, I was sure, this patch will fix 
> the bug reported by Rajkumar S about htt looping. Email No.00333.
> 
> Regards,
> Shao
> Yu Shao wrote:
> 
> > Just a one line patch in case people get infinite loop when start htt 
> > server. It happened to me several times, when iml_if is not set, and 
> > LEBase destructor is called, the code will get Seg fault when try to 
> > deallocate iml_if.
> >
> > Regards,
> > Shao

Regards,
(Continue reading)

Permalink | Reply | 0 comments |
headers
Motonobu Ichimura | 2 Dec 2003 15:53
Favicon

[openi18n-im:00543] Re: A problem in use iml_make_commit_inst in a aux_event

Hi,

On Tue, 2003-12-02 at 19:01, Zarick Lau wrote:
> Hi,
> 
> I have noticed a problem on htt_server! 
> 
> During if_SendEvent () call to le, if the event is IM_EventAux, the
> iml_session_t is a special session (I believe it is a special session
> for XAux communication)
> 
> When I try to call iml function (e.g. commit / env_conv) the call will
> have no effect at all.

Although I didn't check whether it depended on le's code or iiimsf,
this problem was appeared since r11_2 release with
iiimf-skk,iiimf-canna.

In this case, iml function called by IM_EventAux effects XAux Object's
input context, so it looks nothing happend. (Actually, IM events are
propery sent to iiimxcf, but usually XAux Objects don't have an input
area...)

> I have also try to save the last used session (which should a normal
> session/ a valid input context) and calling iml_* with that saved
> session pointer, but 
> again, nothing happens.

> What is the proper way to using iml_* function during a XAux event call?
(Continue reading)

Permalink | Reply | 11 comments |
headers
Hideki Hiura | 2 Dec 2003 21:48

[openi18n-im:00544] Re: a binary package splitting scheme

> From: Yu Shao <yshao <at> redhat.com>
> I am not opposing using "htt", but if htt is the only iiimf
> implementation, iiimf-server would be more friendly to normal users.
> Also few normal users would like to figure out what "xbe" is(I know it
> is easy to distingush it as developer) and then install it, and also
> actually htt-xbe will be invoked by httx, why don't simply put them
> together and call them iiimf-x(or iiimf-htt-x).

Good point. Your idea, such as having consitent iiimf- prefix, follow
user's view point, definitely improves intuitiveness of package name. 

Let's get a closure on naming convention, so it would be consistent
among different distributions.

I summarized the idea currently on the table as below.

I find we agreed on mostly. A few things remaining are:

 1. Should "htt-" be in the server package/the server specific
    component(LEs particular) packages?
 2. Should all lib packages be separated or consolidated into one?
 3. Should the LE packages be categorized per language or name?

My take for 2 is to separate at least for server libs and client libs,
and for 3 is to categorize per LE names, and for 1, either is fine.

Any idea?

Components         | Yu Shao idea   | Roger + Toshi idea | Hideki's suggestion
-------------------+----------------+--------------------+-------------------
(Continue reading)

Permalink | Reply | 0 comments |

Gmane