This is the mail archive of the cygwin mailing list for the Cygwin project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: dlopen() bug (new testcase)

-----Ursprüngliche Nachricht-----
Von: Eric Blake 
Gesendet: Thu, 30 Mar 2006 15:04:27 +0000 (UTC)
>Bernhard Loos <bernloos <at>> writes:
>> >You'll have to read the archives.  Eric Blake has talked about this very
>> >recently.
>> Sorry, but I'm unable to locate it. The last post mentioning IsBad*Ptr 
>> arround March 2005 -  not very recently.
>The gist of it is that IsBad*Ptr is a syscall, and a relatively
>dangerous one at that - Microsoft documents that it is not threadsafe,
>and that calling IsBadWritePtr on a stack guard page will disable
>the ability to grow that stack, meaning that the next time you
>write to the guard page your process will just quit (no popup
>box, no exception, nothing - just quit).  On the other hand, an
>exception handler is faster in the normal case of good pointers,
>does not corrupt system state of whether the stack guard page can
>grow, and does not cause data races with other threads.
>Eric Blake

Thank you very much for explaining this in detail. I have read the other
message bevor, but it doesn't really explain the problems with IsBadReadPtr.

I was able to track the problem down, inserting a call to
_my_tls.init_exception_handler (_cygtls::handle_exceptions);
in the function dll::init() solves it. But I'm unsure, whether it introduce
other bugs. If somebody has checked this, I could submit a patch, if necessary.

                     Bernhard Loos
SMS schreiben mit WEB.DE FreeMail - einfach, schnell und
kostenguenstig. Jetzt gleich testen!

Unsubscribe info:
Problem reports:

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]