[Date Prev][Date Next] [Chronological] [Thread] [Top]

RE: -llunicode (Was: HEADS UP: BerkeleyDB 4.2, libldif, etc.)

I was referring to asymetric encoding and decoding.  IIRC,
one side there was whole struct writes and on the other
side individual fields.  This will fail if there is any
padding in the struct.

I think we should either use a standard encoding (BER,
XDR, etc.) to ensure portability... OR go platform
specific (possibly with swap/size detection but not
correction).  Middle grounds are problematic.

At 04:29 AM 11/26/2003, Howard Chu wrote:
>> -----Original Message-----
>> From: Kurt D. Zeilenga [mailto:Kurt@OpenLDAP.org]
>> At 03:17 PM 11/25/2003, Howard Chu wrote:
>> >It would probably be nice to fix -llunicode's word-size
>> dependencies first
>> >huh?
>> It's worse that word-size issues.  The read/write code is not
>> symmetric... hence the code could possible not read files
>> generated on the SAME platform!
>You're talking about how ucgendat and ucdata don't share a single file header
>I just re-tested HEAD on Solaris sparc64. slapadd was dying in ucdata. After
>I replaced "unsigned long" with "ac_uint4" in ucdata.[ch], ucstr.c, and
>ldap_pvt_uc.h, things ran to completion.
>  -- Howard Chu
>  Chief Architect, Symas Corp.       Director, Highland Sun
>  http://www.symas.com               http://highlandsun.com/hyc
>  Symas: Premier OpenSource Development and Support