OpenSolaris

1. Community: OS/Net (ON)
   1. Blogs
   2. Development Process
   3. Developer's Reference
      1. Introduction
      2. Prerequisites
      3. The Source Tree, part 1
      4. The Source Tree, part 2
      5. Building OpenSolaris
      6. Installing and Testing OpenSolaris
      7. Integration Procedure
      8. Best Practices and Requirements
      9. Glossary
   4. Discussions
   5. Installation (from source) Quickstart
      1. Annotated nightly(1) Mail Example
   6. Currently Known Issues
   7. CRT
      1. Advocates & Sponsors
      2. Becoming a CRT Advocate
      3. Charter
      4. RTI nits to avoid
      5. Sponsor Tasks
      6. Becoming a Sponsor
   8. Schedule
   9. ONNV Flag Days, Heads Ups, and Project Integration History
      1. All
      2. Builds 91-95
      3. Builds 86-90
      4. Builds 81-85
      5. Builds 76-80
      6. Builds 71-75
      7. Builds 66-70
      8. Builds 61-65
      9. Builds 56-60
      10. Builds 51-55
      11. Builds 46-50
      12. Builds 41-45
      13. Builds 36-40
      14. Builds 31-35
      15. Builds 26-30
      16. Builds 21-25
   10. Leaders
   11. Observers
   12. Putback Logs
   13. Developing Solaris
       1. Quality Death Spiral
   14. wx
   15. findunref and unreferenced files
   16. Bourne/Korn Shell Coding Conventions

Heads-up: changes to disassembler in dis, mdb, kmdb

Date: Wed, 31 May 2006 14:22:21 -0700
From: Dan Mick <dan dot mick at sun dot com>
To: onnv-gate at onnv dot eng dot sun dot com
Subject: Heads-up: changes to disassembler in dis, mdb, kmdb

I just integrated a set of changes to libdisasm:

6251453 dis should decode rip-relative memory accesses
6279427 mdb's x64 disassembler doesn't decode %rip-relative addresses for
data access
6427698 mdb/kmdb/dis should look up symbols for immediate operands
6428349 mdb/kmdb/dis (libdisasm) show odd offset for x86 IP-relative
instructions

which changes the x86/x64 disassembly in dis, mdb and kmdb in the following
ways, some of which you'll probably like a lot, but this is just to let you
know that the change is expected:

1) RIP-relative instructions now do symbol lookup, so all the amd64 code
that says things like

main+0xaf:      leaq   0x1f08c1(%rip),%r13
main+0xb6:      movq   0x1f08ba(%rip),%r8

will now report

main+0xaf:      leaq   +0x1f08c1(%rip),%r13     <init_tbl>
main+0xb6:      movq   +0x1f08ba(%rip),%r8      <init_tbl>

yay!  (If there's no symbol available, the full 64-bit effective address is
printed as a hint; this happens for static variables, for instance.)

2) Immediate values will now be looked up and symbols printed; this had
fallen off the shelf somehow in the transition to mdb/kmdb.  (This mostly
affects 32-bit disassembly, as most immediates are rip-relative on amd64.)

3) "Old-style" IP-relative displacements will always print as signed
offsets, and now from "next IP", the way the instruction is actually coded.
So, for example, "jmp" will always have "+/- offset", and "jump to next
instruction" is now disassembled as "jmp +0x0".  (of course symbols are
printed if a lookup succeeds, as always.)

4) Because of 3), the heuristic for "when to print a negative number" has
been relaxed; now an immediate with its sign bit set and absolute value <=
255 is printed as a negative value.  When a negative is printed, the
properly-sized unsigned value is also printed as a hint, in
case that's what the code intended.