[Gs-devel] Catching up on code review

[Igor V. Melichev]

Raph,

> From: gs-devel-admin at ghostscript.com on behalf of Raph Levien
> [raph at acm.org]
> Sent: 10 ьр  2001 у. 20:40
> To: gs-devel at ghostscript.com
> Subject: [Gs-devel] Catching up on code review

> If you have a
> patch that hasn't been responded to, please remind me, so I can
> completely clear out the queue.


Actually I have more 2 things, which a not exactly patches.

1. Long time ago I improved trapezoid algorithm against dropouts.
You reviewed my intermediate version, but after this I completed it,
changing dramatically, and placed to HINTER branch.
I'd like to make a patch for HEAD from this, so as my efforts
go to the main stream of the development. If I'll do soon,
do you have enough time now for working on this ?


2. Applying my old message being unanswered :


- OS memory manager exception when GS HEAD terminates after
  running PDF document "STEUER-RollingMesh 2(radial).ai" on Windows 2000.
  CRTL memory management is corrupted, CRTL incorrectly
  releases memory to OS. Visible only with debugger tool.
  Probably should go to Source Forge bugs.

This annoying thing impedes work on other bugs.
I've analyzed it. The reason is gs_pdf14_device_filter_pop, gdevp14.c, ln
1196 :
it explicitly releases deviceobject, which has 16 references, particularly
from pgs->device.
Few later gs_setdevice_no_init modifies reference counter for the released
device
via pgs->device, so as memory block is modified after it is released.

I am unable to understand why explicit releasing is done here.
Raph, since you worked on this module recently, can you explain this ?

Igor.