From: Michael Glassford (glassfordm_at_[hidden])
Date: 2003-12-04 11:04:26
I'm a bit late responding to this, but I would be interested in seeing the
> Dear Boosters,
> I've experimented a little with the threads library, and soon
> discovered, that it is not possible to use the library in a static
> version on Windows.
> I have read the conversations on this theme in the lists, and
> understand why this is not possible.
> Nevertheless I have written a small lib, that will be helpful to make
> it possible to bind it statically.
> Yet this library is in a state that just demonstrates the principle,
> but in no means can be said to have an optimum interface or beeing
> robust (most error handling code, yet not implemented).
> If there is any interest in the group to share and develop further
> these ideas I would like to contribute the source. Please can anyone
> tell me, if I should simply load up the code to boosts file area? Do
> I need special authorisation for this? I am not even sure if this
> code can ever be tought of beeing
> boost code, since it "solves" only a platform relevant problem.
> A short abstract:
> I am embedding a very small (20k) binary of DLL stub into my library.
> When the process is starting up,
> I copy the binary data to a temporary file (which has flag
> FILE_ATTRIBUTE_TEMPORARY, to
> tell the operating system to keep it in memory, to be fast) and call
> LoadLibrary on this file.
> I end up with a per process DLL (thereby avoiding the 64 Processes
> limit for the task local storage).
> It is straight-forward now to install a callback function, that gets
> called whenever any thread (no matter how it was created) is ending.
> Altough this procedure sounds somewhat tricky, I am using only well
> documented interfaces.
> Best regards,
> Unsubscribe & other changes:
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk