Changes between Initial Version and Version 1 of Ticket #22626, comment 252


Ignore:
Timestamp:
11/19/18 16:00:13 (3 years ago)
Author:
embray
Comment:

Legend:

Unmodified
Added
Removed
Modified
  • Ticket #22626, comment 252

    initial v1  
    88GASMAN needs to know where the bottom of the execution stack is (at least as far as GAP objects are present) in order to properly mark bags that are initialized in global variables, and I believe we may have some cases like that. I think this might explain some weird segfaults and other errors. I'm getting from time to time.  The problem is that GAP assumes that everything stems from an initial `InitializeGap` call that's at the bottom of the stack.  But when using GAP as a library that is clearly not true. 
    99
    10 I'm not sure what the GAP developers intend us to do in this case, as this is clearly a problem for anyone using GAP as a library.  Do we have to manage references ourselves even for local variables?  That would seem fragile, whereas the old `libgap_enter()` very clearly marks the "bottom" of the stack so far as GASMAN needs to be concerned.
     10I'm not sure what the GAP developers intend us to do in this case, as this is clearly a problem for anyone using GAP as a library.  Do we have to manage references ourselves even for local variables?  That would seem fragile, whereas the old `libgap_enter()` very clearly marks the "bottom" of the stack so far as GASMAN needs to be concerned.  I will ask.