| T.R | Title | User | Personal
Name | Date | Lines |
|---|
| 492.1 | can't reproduce the problem | DECWET::JO | Mary had a little lamb, with mint jelly. Dot Warner | Wed Apr 09 1997 13:58 | 13 |
| hi miller,
is it possible for you to upgrade to the RTM version of VC50?
we can't reproduce the problem here using the latest VC50.
we tried it with a DLL built with the RTM kit and couldn't duplicate
the behaviour you describe.
is it possible to get a copy of the DLL along with the linker and rebase
command you're using?
jo
|
| 492.2 | Try Linker option /FIXED:NO | DECWET::MUPOPA | | Wed Apr 09 1997 14:30 | 7 |
| The linker default option is NOW (VC++ 5.0) /FIXED:YES.
The effect is NOT to generate a relocation section in
the DLL or EXE image. Opposite of past behavior.
If you want to allow for relocation, you need to specify
linker option /FIXED:NO.
|
| 492.3 | | HYDRA::CHIN | | Wed Apr 09 1997 14:52 | 7 |
|
Thanks for the information. We are testing the RTM kit as we speak.
Will keep you posted.
Miller
|
| 492.4 | FX!32 ? | nova05.vbo.dec.com::BERGER | | Thu Apr 10 1997 04:20 | 11 |
| > The linker default option is NOW (VC++ 5.0) /FIXED:YES.
>
> The effect is NOT to generate a relocation section in
> the DLL or EXE image. Opposite of past behavior.
Is that true for the Intel version as well ? Isn't that going to
cause problems to FX!32 ? I think I saw somewhere that FX!32 was
based on being able to relocate Intel executables, but I may be
totally wrong...
Vincent
|
| 492.5 | | DECWET::MUPOPA | | Thu Apr 10 1997 17:58 | 5 |
| Ref: .4
The change applies to Intel platforms also.
Yes, this will surprise quite a few people.
I don't know why the change was necessary.
|
| 492.6 | | HYDRA::CHIN | | Wed Apr 16 1997 12:44 | 14 |
|
RE: -1
Sorry for the late reply - the bentley engineer broke his foot last week.
The problem is fixed by changing the default option to /FIXED:NO.
Bentley is able to compile and link for the first pass, and is ready to try
some of those ECO-81 switches. We expect to see some performance gain
against Intel with this build.
Thank you all for the help.
Miller
|
| 492.7 | | TECO::tecotoo.mro.dec.com::mayer | Danny Mayer | Wed Apr 16 1997 16:05 | 6 |
| > Sorry for the late reply - the bentley engineer broke his foot last week.
>
So the fix was a cast??? :-)
Danny
|
| 492.8 | | TALLIS::HERDEG | Mark Herdeg | Wed Apr 23 1997 17:23 | 9 |
| FX!32 does frequently need to relocate x86 images since the FX!32 support images
have to be somewhere in the address space. We were concerned about the VC++ 5.0
change. When I asked about it, I was told that the default has been changed to
/FIXED:YES only for EXEs. It remains /FIXED:NO for DLLs. That is generally OK
with FX!32, unless someone bases their nonrelocatable EXE at an unusual address.
Is that not true? Is the default /FIXED:YES for DLLs as well as EXEs? Yikes!
-Mark
|
| 492.9 | .8 is correct | DECWET::PETERSON | | Thu Apr 24 1997 11:05 | 5 |
| I just built a dll using VC++ V5.0 for intel.
The Base Relocation Directory in it is not empty.
it is possible to specify /fixed with a dll, in which case the Base
Relocation Directory is empty and the DLL cannot be relocated,
but the default is to have relocations with a DLL
|