60823 - UnicodeToGBK and UnicodeToGB2312 converter don't work on SPARC Solaris

Status: VERIFIED FIXED

(Core :: Internationalization, defect, P1)

Description

[Masaki Katakai]

nsUnicodeToGBK.cpp and nsUnicodeToGB2312V2.cpp don't work
for ASCII characters on SPARC Solaris. Please try the following,

 ./nsconv -f utf-8 -t x-gbk

When you type ascii characters, but it doesn't return.

On SPARC Solaris platform, Mozilla could not create
new profile under user's home directory in zh.GBK locale
but created MOZILLA_HOME/default_prefs.js.

After debugging, I found NS_NewUnicodeLocalFile() in zh.GBK locale.

This will convert UNICODE to native encoding. The target
string is very simple, e.g. "/home/katakai/.mozilla" but
it fails and returns "".

It seems that nsUnicodeToGBK.cpp and nsUnicodeToGB2312V2.cpp
uses leftbyte when the input is ASCII, but it doesn't work
on SPARC platform.

I'll attach the simple patch for this problem. Please evaluate.

Comment 1

[Masaki Katakai]

Attachment: changes for nsUnicodeToGB2312V2.cpp and nsUnicodeToGBK.cpp

Comment 2

[nhottanscp]

Katakai san, I assign this to you since you have a patch, cc to erik, bstell.

Comment 3

[Erik van der Poel]

The fix is correct, but I would prefer the following:

  *aDest++ = *pSrc;

r=erik

Comment 4

[Masaki Katakai]

Hotta-san, can you check-in this change for me?
I've requested CVS access but I could not get yet.

Comment 5

[nhottanscp]

Please get a super review then I can check in.

http://www.mozilla.org/hacking/reviewers.html

Comment 6

[tao]

Hi, Masaki-san:

Please resubmit a patch w/ Erik's suggestion and get a review from cata and
super-review from Erik.

thx

Comment 7

[Masaki Katakai]

Attachment: revised patch

Comment 8

[tao]

Hi, Erik and Naoki:


Would you review (sr) the proposed patch ?



thx

Comment 9

[Erik van der Poel]

sr=erik

Comment 10

[nhottanscp]

Why *aDest = pSrcDBCode->leftbyte; does not work on SPARC?

Comment 11

[Erik van der Poel]

SPARC is big-endian, while Intel is little-endian. The data is in little-endian
double-byte Unicode (shorts), but the code is applying a struct that has
leftbyte first, so it doesn't work.

Comment 12

[Erik van der Poel]

Sorry, what I meant to say, was: The data is in little-endian Unicode only on
Intel, so the struct with leftbyte will work there, but the data is big-endian
on SPARC, so the struct does not work there.

Comment 13

[Shanjian Li]

I found the same problem on HP (bug 61466). I will mark 61466 as a duplicated one.
Katakai, please use type cast before assignment, that will avoid some compiler 
warning. ie.  *aDest++ = (char)*pSrc;
If you can check it into trunk, please do it. Otherwise reassign it to me.
thanks!

Comment 14

[nhottanscp]

r=nhotta, please include shanjian's comment

Comment 15

[Masaki Katakai]

Attachment: added (char) cast, Shanjian, could you check-in for me?

Comment 16

[Masaki Katakai]

I couldn't still get a permission for check-in. Reassign to Shanjian.

Comment 17

[Shanjian Li]

*** Bug 61466 has been marked as a duplicate of this bug. ***

Comment 18

[tao]

Hi, Katakai-san:

We got r=nhotta & sr=erik for the proposed patch. Is there anything else
to be done?

Comment 19

[Shanjian Li]

Fix has been checked in to trunk and OEM branch. 

Comment 20

[Masaki Katakai]

integrated into OEM branch. Tao, Shanjian, thank you very much for your help.

Comment 21

[Teruko Kobayashi]

Changed QA Contact to [e-mail].

Comment 22

[Masaki Katakai]

Verified fix on the latest build on Solaris SPARC.

Comment 23

[Linda Baliman]

Adding intl and nsbeta1 keywords.

Adding mcarlson to cc-list.