697442 - "error: cannot apply 'offsetof' to a non constant address" with gcc 4.6

Status: RESOLVED WONTFIX

(Tamarin Graveyard :: Build Config, defect)

Description

[Nav]

User Agent: Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/535.2 (KHTML, like Gecko) Chrome/15.0.874.102 Safari/535.2

Steps to reproduce:

Synced to a0d75bba6a06.
Built avmshell.


Actual results:

core/CodegenLIR.cpp:4522:69: error: cannot apply 'offsetof' to a non constant address



Expected results:

Build without errors.

Comment 1

[Felix S. Klock II [:pnkfelix, :fklock]]

A bit more info, from:
  http://asteam.corp.adobe.com/irc/log/ttlogger/tamarin/20111026

  <fklockii> (This problem seems like it can get into some very twisty 
             passages. The fact that it claims that
             offsetof(ObjectVectorObject, m_list.m_data) is a 
             non-constant address strikes me as strange, *unless* its 
             doing something funky with how it compiling templates. But 
             maybe I am missing something.)
  ...
  <fklockii> Try going back to say maybe rev 6531. Thats from right
             before Lars started putting vector optimzations into
             CodegenLir.
  <nav>      fklockii: I can build 6531 without problem.
  <fklockii> nav: great to hear. So it sounds like we're just getting 
             unlucky in that we're trying to be clever in two directions
             at once.

In particular, I suspect the recently added vector JIT optimizations, such as the ones added with Bug 599099, have code that contains fragile uses of offsetof.

Comment 2

[Sylvestre Ledru [:Sylvestre]]

Tamarin is a dead project now. Mass WONTFIX.

Comment 3

[Sylvestre Ledru [:Sylvestre]]

Tamarin isn't maintained anymore. WONTFIX remaining bugs.