Hello Jeroen,
> Since .DEF files are an MS invention, it's likely the Borland
> linker just ignores directives it doesn't know to increase
> compatibility.
Something I thought of as well. But than again, what good is
"compatibility" when the results are not the same. :-\
> Pretty sure there isn't, since it's redundant with explicitly
> specifying ordinals.
Quite so. On the other hand, 1) starting named exports at a certain offset
(without explicitily defining ordinals for all of them) than becomes a
possibility 2) as the linker does not do it automatically I assumed that its
needed to be done explicitily.
> The case of "I want some functions to appear at these
> ordinals, but please automatically do something sensible
>for the rest" probably wasn't considered by the toolsmiths.
That it actually does: The ordinal export went where I wanted it, the named
export was put at ordinal 1 (the first free one).
The situation I asked the question about was a single ordinal export at (for
instance) 10. A dump of the DLL showed me 10 exported adresses, 1 used.
Thats, in the DLL, wasting 9 entries in two tables.
> Worst case you'll have to write a tool yourself to generate the
> .DEF file with ordinals like you want them.
:-) Yes, that won't be a problem. Its just that I see non-optimal
behaviour and like to know if I can alter it.
The problem how to let the linker generate an ordinal-only export in the DLL
remains though. Oh well, if push comes to shove I can always write
something which will patch the generated DLL (using either the DEF file or
some added info in the sourcefile), removing the names from the apropriate
tables.
Thanks for the answers.
Rudy Wieser
-- Origional message
Jeroen Mostert <
jmos...@xs4all.nl> schreef in berichtnieuws
513dbc6e$0$6909$e4fe...@news2.news.xs4all.nl...