Rewrite of IBM doublebyte charsets (original) (raw)

Ulf Zibis Ulf.Zibis at gmx.de
Thu May 21 20:11:13 UTC 2009

Am 21.05.2009 00:22, Xueming Shen schrieb:

Ulf, thanks for the review!

Now we have a bugid for this one.

Good idea! I set it on my watch list.

With this change, (1)21% reduce of rt.jar size (from 2.99M-with the euctw reduce to 2.35M)

??? My rt.jar is 46 MB. Can you explain your numbers?

The test case test/sun/nio/cs/OLD/TestIBMDB.java is used to make sure there is no mapping regression, there are new mapping entries added (because switching to the latest IBM cdc mapping tables for some charsets, which the b2c tables we have are not accurate/complete), but no conflit with previous mappings.

In IBM933.map you state:

Warning:

(2) we "should" have an entry

25 000a

in IBM933.nr (b->c only tables) as other ebcdic

mappings do, but the "old" implementation actually

maps \u000a to 25. Keep it old behavior for now.

I think we shouldn't stick on old behaviour here, as there obviously was an error in old code, but nobody filed a bug until now.

Alan, Martin, I need you guys also help review the change, I will need a name on the putback message.

I would be happy to see my name in 2nd/3rd position. :-)

-Ulf

More information about the core-libs-dev mailing list