#5292: libHSghc exports more symbols than Windows can handle
---------------------------------+------------------------------------------
Reporter: batterseapower | Owner:
Type: bug | Status: new
Priority: normal | Component: Compiler
Version: 7.0.3 | Keywords:
Testcase: | Blockedby:
Os: Windows | Blocking:
Architecture: Unknown/Multiple | Failure: Runtime crash
---------------------------------+------------------------------------------
If you build GHC with shared libraries enables, the generated libHSghc.dll
will export 65536 symbols. This is the limit that Windows DLLs can handle
(the source of the limitation is that they use 16-bit integers to
represent "ordinals").
Unfortunately, this means that the generated DLL is actually missing a ton
of exports! Bad things happen at this point, including runtime crashes in
GHCs linker. One interesting feature is that there seems to be a bug with
GetProcAddress for modules that have filled up their export lists like
this, so if hmod is a handle to the loaded DLL, GetProcAddress(hmod,
"LocalAlloc") will succeed even though LocalAlloc is not exported by the
DLL.
The number of symbols exported varies wildly depending on how you build
the compiler. A "quick" build (from build.mk.sample) only exports 39k
symbols, which is under the limit, but the kind of build that validate
does ends up hitting the limits. (This difference was the source of some
fun debugging confusion for me...)
Because of this bug, we won't be able to distribute a dynamically linked
GHC on Windows. It is also preventing me from getting plugins06 to pass on
Windows.
One workaround would be to build GHC with settings that result in less
exported symbols until we get below the limit again -- i.e. export less
unfoldings. This might adversely impact compiler performance, however.
--
Ticket URL: <http://hackage.haskell.org/trac/ghc/ticket/5292>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler
_______________________________________________
Glasgow-haskell-bugs mailing list
[email protected]
http://www.haskell.org/mailman/listinfo/glasgow-haskell-bugs
