Minheadless trial
Locked
 
|
wow time flies, its been 18 months since Ronie announced his Minheadless VM branch... [1] http://forum.world.st/Minheadless-VM-flavour-status-update-td4928091.html I've been meaning this whole time to try it out. Recently in my meandering thoughts on FreeRTOS port, I guessed Minheadless might make a good starting basis to experiment with. So a quick test on Ubuntu16.04.3(Windows10-WSL)... Long ago I'd already done... git clone [hidden email]:OpenSmalltalk/opensmalltalk-vm.git so I now did... cd opensmalltalk-vm git remote add ronie [hidden email]:ronsaldo/opensmalltalk-vm.git git fetch ronie git checkout MinimalisticHeadless ./scripts/updateSCCSVersions cd build.minheadless.cmake/x64/pharo.stack.spur/ ./mvm VM=`pwd`/release/dist/pharo ls $VM ==> confirmed it exists cd ~/pharo wget http://files.pharo.org/image/70/latest-minimal-64.zip unzip latest-minimal-64.zip sudo $VM Pharo7.0-metacello-64bit-65cff7b.image eval "19 + 23" ==> 42 yay Particularly interesting from [1] is "using the same CMake scripts for building on the three platforms: Windows, Linux and OS X." So I thought I'd also give it a try under Windows10-VisualStudio2017. I hardly used VS before, so I'll detail what I did as a reminder for myself (and also in case someone might be able to help if I get stuck). Under a windows cmd.exe prompt (having installed git and cloned opensmalltalk-vm a long time ago) cd opensmalltalk-vm git remote add ronie [hidden email]:ronsaldo/opensmalltalk-vm.git git fetch ronie git checkout -b MinimalisticHeadless ronie/MinimalisticHeadless I watched video [2] about CMake integration with Visual Studio 10. and checked the VS workloads configuration specified [2@1:13] using [3]. There is a CMakeLists.txt file in the git root folder, so.. started VS > File > Open > Folder... C:\Repos\OpenSmalltalk\opensmalltalk-vm The CMake command run ([2]@2:13) was... C:\...\cmake.exe -G "Ninja" -DCMAKE_INSTALL_PREFIX:PATH="C:\Users\Ben\CMakeBuilds\...\x86-Debug" -DCMAKE_CXX_COMPILER="C:/.../MSVC/14.14.26428/bin/HostX86/x86/cl.exe" -DCMAKE_C_COMPILER="C:/.../MSVC/14.14.26428/bin/HostX86/x86/cl.exe" -DCMAKE_BUILD_TYPE="Debug" -DCMAKE_MAKE_PROGRAM="C:\...\CMAKE\Ninja\ninja.exe" "C:\Repos\OpenSmalltalk\opensmalltalk-vm" I changed the cmake configuration [2@2:27] to "x64-Debug". For debug target [2@3:00] I chose "pharo.exe" Hey, thats pretty cool [2@4:05] that switching between targets changes the highlighting of #ifdef code. I had a moment when I was missing the "Debug and Launch Settings" menu item [2@4:40-4:55] and mentioned at [4], but maybe there was some clash with previously experimenting with git cloning opensmalltalk-vm from inside VS. After closing VS and reopening my cmd.exe git'd opensmalltalk-vm it showed up, but I didn't change any settings there. Then (with x64-Debug & pharo.exe selected) I went Cmake > Build All ==> 10 Errors, 9 Warnings, which I'll discuss in a followup post. cheers -ben [2] https://blogs.msdn.microsoft.com/vcblog/2016/10/05/cmake-support-in-visual-studio/ [3] https://docs.microsoft.com/en-us/visualstudio/install/modify-visual-studio [4] https://blogs.msdn.microsoft.com/vcblog/2016/12/20/cmake-support-in-visual-studio-2017-whats-new-in-the-rc-update/ [5] https://developercommunity.visualstudio.com/content/problem/234808/cmake-debug-and-launch-settings-not-shown-while-bu.html |
|
|
On 6 August 2018 at 11:50, Ben Coman <[hidden email]> wrote: > wow time flies, its been 18 months since Ronie announced his > Minheadless VM branch... > [1] http://forum.world.st/Minheadless-VM-flavour-status-update-td4928091.html In Visual Studio... > (with x64-Debug & pharo.exe selected) > I went Cmake > Build All > ==> 10 Errors, 9 Warnings, which I'll discuss in a followup post. and btw, the build finishes too quick Here are the errors... ------------------------ https://github.com/ronsaldo/opensmalltalk-vm/blob/be7b1c03/platforms/minheadless/windows/sqPlatformSpecific-Win32.c#L80 typedef HRESULT WINAPI (*SetProcessDpiAwarenessFunctionPointer) (int awareness); C2059 sqPlatformSpecific-Win32.c:80 syntax error: '(' E0651 a calling convention may not be followed by a nested declarator. The following change reduces build errors to 1... typedef HRESULT (*SetProcessDpiAwarenessFunctionPointer) (int awareness); but I'm not sure of the implications. Could be "the parenthesis around the function name" [2] Or "there's already a C standard function [of that name]" [3] The Microsoft definition returns a BOOL [4]. [2] https://software.intel.com/en-us/forums/intel-c-compiler/topic/385282 [3] https://social.msdn.microsoft.com/Forums/vstudio/en-US/2c57ada8-e1c9-4030-a058-6a7afd2fd29f/a-calling-convention-may-not-be-followed-by-a-nested-declarator?forum=vcgeneral [4] https://docs.microsoft.com/en-us/windows/desktop/api/winuser/nf-winuser-setprocessdpiawarenesscontext ------------------------ https://github.com/ronsaldo/opensmalltalk-vm/blob/be7b1c03/platforms/minheadless/common/sqWindow-Dispatch.c#L68 sqInt ioSetCursorARGB(sqInt cursorBitsIndex, sqInt extentX, sqInt extentY, sqInt offsetX, sqInt offsetY) C2371 'ioSetCursorARGB': redefinition; different basic types where sqInt is "long long". but the function declaration is https://github.com/ronsaldo/opensmalltalk-vm/blob/21cc02b6bf/platforms/minheadless/windows/sqPlatformSpecific-Win32.h#L117 int ioSetCursorARGB(sqInt bitsIndex, sqInt w, sqInt h, sqInt x, sqInt y); After modifying that function declaration to return 'sqInt', "Cmake > Build all" runs much longer ------------------------ and ends up with 4 link errors (and 6600 warnings) Pharo.exe LNK1120 - 1 unresolved external Pharow.exe LNK1120 - 1unresolved external sqMain.c.obj LNK2019 - unresolved external symbol _imp_osvm_main referenced in function main sqWin32Main.c.obj LNK2019 - unresolved external symbol _imp_osvm_main referenced in function WinMain I do see osvm_main is defined here... https://github.com/ronsaldo/opensmalltalk-vm/blob/be7b1c03/platforms/minheadless/common/sqVirtualMachineInterface.c#L618 but I'm stuck, I don't understand why its looking for "_imp_osvm_main" I'd be glad if anyone can advise on this, and the two errors above. cheers -ben |
|
|
On 6 August 2018 at 13:22, Ben Coman <[hidden email]> wrote: > and ends up with 4 link errors (and 6600 warnings) > Pharo.exe LNK1120 - 1 unresolved external > Pharow.exe LNK1120 - 1unresolved external > sqMain.c.obj LNK2019 - unresolved external symbol _imp_osvm_main referenced in function main > sqWin32Main.c.obj LNK2019 - unresolved external symbol _imp_osvm_main referenced in function WinMain > > I do see osvm_main is defined here... > https://github.com/ronsaldo/ > > but I'm stuck, I don't understand why its looking for "_imp_osvm_main" I see that sqMain.c (https://github.com/ronsaldo/opensmalltalk-vm/blob/be7b1c03/platforms/minheadless/common/sqMain.c) only defines a single function... #include "OpenSmalltalkVM.h" int main(int argc, const char **argv) { return osvm_main(argc, argv); } To check that the osvm_main function is available to be linked, at the end of CmakeLists.txt I see... # Build the VM executable(s) add_executable(${VM_EXECUTABLE_NAME} platforms/minheadless/common/sqMain.c) target_link_libraries(${VM_EXECUTABLE_NAME} ${VM_LIBRARY_NAME} ${VM_DEPENDENCIES_LIBRARIES}) Using https://stackoverflow.com/questions/9298278/cmake-print-out-all-accessible-variables-in-a-script I found... VM_LIBRARY_NAME=PharoVMCore where that library is defined VM_CORE_LIBRARY_TYPE=STATIC add_library(${VM_LIBRARY_NAME} ${VM_CORE_LIBRARY_TYPE} ${VM_SOURCES} ${VM_INTERNAL_PLUGIN_SOURCES}) and an inspection of PharoVMCore.lib (using 7zip to expand it) shows it contains function osvm_main symbol, but not the "_imp_" prefixed symbol. Regarding the "_imp_" prefix, the next to last comment in this thread gave a hint about the __declspec() needing to be dllexport instead of dllimport. The declaration for osvm_main() in OpenSmalltalkVM.h is effectively... # define OSVM_VM_EXPORT __declspec(dllexport) # define OSVM_VM_IMPORT __declspec(dllimport) #ifdef BUILD_VM_CORE # define OSVM_VM_CORE_PUBLIC OSVM_VM_EXPORT #else # define OSVM_VM_CORE_PUBLIC OSVM_VM_IMPORT #endif OSVM_VM_CORE_PUBLIC OSVMError osvm_main(int argc, const char **argv); So the following change to sqMain.c fixes the link error... #define BUILD_VM_CORE #include "OpenSmalltalkVM.h" int main(int argc, const char **argv) { return osvm_main(argc, argv); } and similar for sqWin32Main.c. |
|
|
On 6 August 2018 at 23:54, Ben Coman <[hidden email]> wrote:
The next roadbump is... LNK2019 unresolved external symbol 'alloca' referenced in function callIA32IntegralReturn -- File PharoVMCore.lib(xabicc.c.obj) From the root folder, doing... find . -name xabicc.c result in... ./platforms/Cross/plugins/IA32ABI/xabicc.c Googling for "alloca Visual Studio 2017" lead me to... which indicates there is no "alloca" only "_alloca" Now "
xabicc.c" is a stub for including platform specific C code, and the lovely defines-highlighting of Visual Studio indicates "x64win64abicc.c" is the active one right now, so as a quick hack,
jamming a define just above that include... #if i386|i486|i586|i686 # include "ia32abicc.c" #elif powerpc|ppc # include "ppc32abicc.c" #elif x86_64|x64|__x86_64|__x86_64__|_M_AMD64|_M_X64 # if _WIN64 # define alloca _alloca // <<<QUICK HACK
# include "x64win64abicc.c" # else # include "x64sysvabicc.c" # endif #elif __ARM_ARCH__|__arm__|__arm32__|ARM32 # include "arm32abicc.c" #endif allows CMake > Build All ==> Build succeeded. Yay! To determine a suitable permanent home for that define, doing... grep -R " _alloca" *
from the root folder shows... platforms/Cross/plugins/IA32ABI/arm32abicc.c:# define alloca _alloca platforms/Cross/plugins/IA32ABI/ia32abicc.c:# define alloca _alloca platforms/Cross/plugins/IA32ABI/ppc32abicc.c:# define alloca _alloca platforms/Cross/plugins/SqueakFFIPrims/sqFFIPlugin.c:# define alloca _alloca platforms/minheadless/windows/sqPlatformSpecific-Win32.h:# define alloca _alloca platforms/unix/config/configure:# define alloca _alloca platforms/win32/vm/sqPlatformSpecific.h:# define alloca _alloca processors/ARM/gdb-7.10/include/alloca-conf.h:# define alloca _alloca processors/IA32/bochs/bx_debug/parser.c:# define alloca _alloca src/plugins/SqueakFFIPrims/ARM32FFIPlugin.c:# define alloca _alloca src/plugins/SqueakFFIPrims/IA32FFIPlugin.c:# define alloca _alloca src/plugins/SqueakFFIPrims/X64SysVFFIPlugin.c:# define alloca _alloca src/plugins/SqueakFFIPrims/X64Win64FFIPlugin.c:# define alloca _alloca Interesting that there is a similar define for ia32abicc.c, so comparing that with x64win64abicc.c (at https://www.diffchecker.com/mqCWMaTV) looks like it should be slotted into the latter about line 41. Doing that ==> Build successful. cheers -ben |
|
|
Hi Ben,
On Mon, Aug 6, 2018 at 12:57 PM, Ben Coman <[hidden email]> wrote:
Feel free to make this edit and commit
_,,,^..^,,,_ best, Eliot |
|
|
In reply to this post by Ben Coman
On 7 August 2018 at 05:12, Eliot Miranda <[hidden email]> wrote:
I'm pushing changes here... and the diff can be tracked here... ------------------------ On 6 August 2018 at 13:22, Ben Coman <[hidden email]> wrote: On 6 August 2018 at 11:50, Ben Coman <[hidden email]> wrote: I found the correct solution to this... "The trick is placing the [call declaration] inside the parentheses" i.e. the following compiles cleanly
typedef HRESULT (WINAPI * ----------------------------- Now running the VM (without parameters) I get... Debug Assertion Failed! Program: ...\x64-Debug\dist\pharo.exe File: minkernel\crts\ucrt\src\appcrt\tran\amd64\ieee.c Line: 106 Expression: (mask&~(_MCW_DN | _MCW_EM | _MCW_RC))==0 at the call to _controlfp(FPU_DEFAULT, _MCW_EM | _MCW_RC | _MCW_PC | _MCW_IC); x64 does not support _MCW_PC or _MCW_IC but I'm clueless about the implications of these FPU flags. Could our math guys please advise? Eliminating those two flags allows a VM run successfully without loading an Image. i.e. it successfully passes... osvm_initialize(); osvm_parseCommandLineArguments(argc, argv); osvm_initializeVM(); Next is to try loading an Image. cheers -ben |
|
|
On 7 August 2018 at 13:22, Ben Coman <[hidden email]> wrote:
+ Downloaded and unzipped
http://files.pharo.org/image/7 + In Visual Studio, right-clicked CMakeLists.txt > Debug and Launch Settings > pharo.exe + Edited the presented "launch.vs.json" as follows... { "version": "0.2.1", "defaults": {}, "configurations": [ { "type": "default", "project": "CMakeLists.txt", "projectTarget": "pharo.exe (dist\\pharo.exe)", "name": "pharo.exe with image (dist\\pharo.exe)", "args": [ "C:\\#Dev\\latest-minimal-64\\Pharo7.0-metacello-64bit-65cff7b.image", "s" ] }, ] } Now running "x64-Debug" and "pharo.exe with image" I find it gets hung up in... sqAllocateMemory(usqInt minHeapSize, usqInt desiredHeapSize) in call... alloc = sqAllocateMemorySegmentOfSizeAboveAllocatedSizeInto (roundUpToPage(desiredHeapSize), address, &allocBytes); with... desiredHeapSize ==> 39583744 (~39MB) address ==> 0x00007ff7d6e00000 where inside that call... void *sqAllocateMemorySegmentOfSizeAboveAllocatedSizeInto(sqInt size, void *minAddress, sqInt *allocatedSizePointer)" has... size ==> 39583744 minAddress ==> 0x00007ff7d6e00000 address ==> 0x00000000d6e00000 bytes ==> 39583744
(~39MB)
delta ==> 1048576 it gets stuck the following loop... https://github.com/bencoman/opensmalltalk-vm/blob/8231b96f9b/platforms/win32/vm/sqWin32SpurAlloc.c#L139-L183 while ((usqIntptr_t)(address + bytes) > (usqIntptr_t)address) { alloc = VirtualAlloc(address, bytes, MEM_COMMIT|MEM_RESERVE, PAGE_READWRITE); ... if (alloc >= (char *)minAddress) && (alloc <= address + delta)) return alloc; ... address += delta; } where... alloc ==> 0x00000000d6e00000 (alloc >= (char *)minAddress) ==> false. which is effectively... (does this seem normal?) ((char *)minAddress-address)/delta ==> 134,180,864 loops before exit. but... (alloc >= (char *)address) ==> true. would exit on the first pass of the loop, except... /* For some reason (large page support?) we can ask for a page-aligned * address such as 0xNNNNf000 but VirtualAlloc will answer 0xNNNN0000. * So accept allocs above minAddress rather than allocs above address */ where back one level "minAddress" is derived like this... /* choose a suitable starting point. In MinGW the malloc heap is below the * program, so take the max of a malloc and something form uninitialized * data. */ hint = malloc(1); free(hint); hint = max(hint,(char *)&fIsConsole);
address = (char *)(((usqInt)hint + alignment - 1) & ~(alignment - 1)); address passed-by-value into minAddress Now I'm out of my depth trying to understand the background on that logic, but if I blindly change the loop exit condition from...
(alloc >= (char *)minAddress)
to...
(alloc >= (char *)address)
then osvm_loadImage(tempImageNameAttempt() and subsequently osvm_loadDefaultImage() succeed. So some expert attention would be good here (as time permits) In the meantime, I'll forge ahead to play with osvm_run(); cheers -ben P.S. making "#define printProbes 1" and "fIsConsole = 1;" shows... probing [00000000D6E00000,00000000D93C0000) probing [00000000D6F00000,00000000D94C0000) probing [00000000D7000000,00000000D95C0000) probing [00000000D7100000,00000000D96C0000) probing [00000000D7200000,00000000D97C0000) probing [00000000D7300000,00000000D98C0000) probing [00000000D7400000,00000000D99C0000) probing [00000000D7500000,00000000D9AC0000) probing [00000000D7600000,00000000D9BC0000) |
|
|
In reply to this post by Ben Coman
On 7 August 2018 at 13:47, Esteban Lorenzano <[hidden email]> wrote:
I'll have a look at it.
That is good to know. So just to be clear... even though its name indicates "headless", you are running a full graphical Image? I got more curious about the original announcement saying... "minheadless is a VM variant that unifies a lot of the code for Windows, Linux and OS X." So here https://docs.google.com/spreadsheets/d/183GnSKZSMSEgVQZ0Fqa7-vITqKSpAlW_AxnZKl7x048/edit?usp=sharing I compare... find "platforms/Win32/vm" -exec wc {} \;
find "platforms/Unix/vm" -exec wc {} \;
find "platforms/Mac OS/vm" -exec wc {} \; to...
find "platforms/minheadless" -exec wc {} \; and if nothing important has been missed it seems about 30% of the original size. I'd expect Ronie's effort to produce this should have a massive impact on future maintainability. cheers -ben P.S. I excluded the following presuming its common before/after. find "platforms/Cross/vm" -exec wc {} \; btw, its about half the size of platforms/minheadless. |
|
|
In reply to this post by Ben Coman
On 7 August 2018 at 19:47, Ben Coman <[hidden email]> wrote: > > it gets stuck the following loop... https://github.com/bencoman/ whoops, that was meant to be the minheadless file....
I've hardly used Visual Studio before, and never combined with CMake, so I'll record a few things I discover in case others might find it interesting or advise a better way. To try with the Stack Interpreter first, in CMakeLists.txt changed... option(COG_JIT "Cog JIT" ON) to... option(COG_JIT "Cog JIT" OFF) then cleared the cache... Right-clicked on "CMakeLists.txt > Cache(x64-Debug only) > Delete cache folder..." otherwise it didn't seem to pick up that change. To give the VM an image to run and something to do... Right-clicked on "CMakeLists.tx then edited the presented "launch.vs.json" as follows... { "version": "0.2.1", "defaults": {}, "configurations": [ { "type": "default", "project": "CMakeLists.txt", "projectTarget": "pharo.exe (dist\\pharo.exe)", "name": "pharo.exe minImage+eval (dist\\pharo.exe)", "args": [ "C:\\#Dev\\latest-minimal-64\\ ] }, ] } Then... "CMake > Build All" followed by starting... "x64-Debug" with "pharo.exe minImage+eval" which opened a console window that doesn't do much besides output like this...
GIV(remapBufferCount) == 0 47923
Now I notice the file "dist\stderr" says... "Your VM is too old for this image. Please download the latest VM." so probably I need to run "scripts\updateSCCSVersions" but thats a shell script. How to run it under Windows?
Luckily my installed Git Windows has Git Bash which opens a cygwin environment. where I do... cd /c/#Repos/OpenSmalltalk/opensmalltalk-vm/ ./scripts/ updateSCCSVersions
Now to give a bit more visibility to program execution, in "spurstack64src/interp.c" I inserted the following into main interpreter loop... "+" unsigned long long debugCountBytecode = 0; while (1) { "+" debugCountBytecode += 1; "+" if ((debugCountBytecode % 100000 == 0)) "+" { "+" printf("%llu\n", debugCountBytecode); "+" } Then... "CMake > Build All" followed by starting... "x64-Debug" with "pharo.exe minImage+eval" which opened a console window with the following output... GIV(remapBufferCount) == 0 47923 GIV(remapBufferCount) == 0 42469 100000 200000 300000 400000 500000 600000 700000 800000 900000 1000000 1100000 1200000 1300000 1400000 1500000 and updated contents of these files... yay! dist\stderr ==> #'EvaluateCommandLineHandler successfully finished' dist\stdout ==> 42 However the VM/Image keeps running.
Adding "--quit" to the command line didn't help.
I'm not really familiar with the command line handlers.
Doing... "Debug > Break All"
bring me up some warnings like... "winmm.pdb not loaded" "winmm.pdb contains the debug information required to find the source for the module winmm.dll" Never heard of a PDB, but loaded them by following these instructions... "Specify Symbol (.pdb) and Source Files in the Visual Studio Debugger" and learnt a lot more about them... "PDB Files: What Every Developer Must Know" Now the next few times I started...
"x64-Debug" with "pharo.exe minImage+eval"
it got hung up downloading the symbols. I had to terminate & restart debugging it a few times as it progressed a bit each time. When finally got running again and after
1500000 printed I did...
"Debug > Break All"
and... "Debug > Windows > Threads" and now it shows the proper names rather than just addresses...  A little surprised by the number of threads. No clue what ntdll.dll!NtWaitForWorkViaWorkerFactory or winmm.dll!timeThread are, but leave that for later.  invoked from ioRelinquishProcessorForMicroseconds() and beatThreadStateMachine(). Clicking on each of those displays their source code indicating the "external code" for each is MsgWaitForMultipleObjects() and WaitForSingleObject()
respectively
. Its very easy to flip back and forth between the two application threads, stepping-out of the "external code" with Shift-F11, to observe a tiny window of VM operation. Well thats enough for one post! Success running a MSVC compiled Spur Stack "minheadless" VM. Next I'll try...
option(COG_JIT "Cog JIT" ON)
cheers -ben P.S. Definitions for the curious... sqInt ioRelinquishProcessorForMicroseconds(sqInt microSeconds) { /* wake us up if something happens */ ResetEvent(vmWakeUpEvent); MsgWaitForMultipleObjects(1, &vmWakeUpEvent, FALSE, microSeconds / 1000, QS_ALLINPUT); ioProcessEvents(); /* keep up with mouse moves etc. */ return microSeconds; } static DWORD WINAPI beatThreadStateMachine(void *careLess) { beatThreadState = active; while (beatThreadState != condemned) { DWORD res = WaitForSingleObject(beatSemaphore, beatThreadState == active ? beatMilliseconds : INFINITE); if (res == WAIT_TIMEOUT) heartbeat(); else if (res == WAIT_FAILED) abortMessage(TEXT("Fatal: WaitForSingleObject(beatSemaphore) %ld"), GetLastError()); else printLastError(TEXT("beatThreadStateMachine WaitForSingleObject")); } beatThreadState = dead; return 0; } |
|
|
Hi Ronie, Esteban, Could you do a quick review of my CMakeLists.txt change for any obvious blunder. This may be my last post on this thread for a short while. On 8 August 2018 at 21:39, Ben Coman <[hidden email]> wrote:
I did that, and then... "CMakeLists.txt > Cache(x64-Debug only) > Delete cache folder..." "CMake > Build All" then started "x64-Debug" , "pharo.exe minImage+eval" and a console window opens that reports... "linking callsite to invalid address"Well I'm not sure how ready is the Windows 64bit JIT is, so lets try with 32bit JIT. In VS selecting "x86-Debug" , "pharo.exe minImage+eval" and "Cmake > Build All" reports... "Severity Code Description Project File Line Suppression State Error C1189 #error: As yet no Cogit implementation appears to exist for your platform. C:\Program Files (x86)\Microsoft Visual Studio\2017\Community\Common7\ In the CMake variables SourceFolderName=spur64src because..
VM_64BITS=TRUE which is set here... if(SQUEAK_PLATFORM_X86_64) if(BUILD_I386_VERSION) set(SQUEAK_PLATFORM_X86_32 True) set(SQUEAK_PLATFORM_X86_64 False) else() set(VM_64BITS TRUE) set(SourceFolderName "${SourceFolderName}64") if(WIN32) set(FFI_VARIANT_X64_WIN64 True) else() set(FFI_VARIANT_X64_SYSV True) endif() set(VM_TARGET_CPU "x86_64") endif() else() set(VM_64BITS False) endif() which I'm not clear how that logic works, since it seems BUILD_I386_VERSION doesn't prevent VM_64BITS=True Rearranging like... if(BUILD_I386_VERSION) set(SQUEAK_PLATFORM_X86_32 True) set(SQUEAK_PLATFORM_X86_64 False) set(VM_64BITS False) else() endif() if(SQUEAK_PLATFORM_X86_64) set(VM_64BITS TRUE) set(SourceFolderName "${SourceFolderName}64") if(WIN32) set(FFI_VARIANT_X64_WIN64 True) else() set(FFI_VARIANT_X64_SYSV True) endif() set(VM_TARGET_CPU "x86_64") else() endif() then adding
-DBUILD_I386_VERSION in
"CmakeSettings.json" like this... "CMakeLists.txt > Change Cmake Settings" { "name": "x86-Debug", "generator": "Ninja", "configurationType": "Debug", "inheritEnvironments": [ "msvc_x86" ], "buildRoot": "${env.USERPROFILE}\\ "installRoot": "${env.USERPROFILE}\\ "cmakeCommandArgs": "-DBUILD_I386_VERSION=ON", "buildCommandArgs": "-v", "ctestCommandArgs": "" }, then refreshing... "CMakeLists.txt > Cache(x64-Debug only) > Delete cache folder..." "CMake > Build All"
leaves these values in the Cmake variables... VM_64BITS=False SourceFolderName=spursrc and the build completes successfully. I'm leaving it there for the moment. Actually the above sat in unchanged in draft for a week since I've had to shift my attention to something else. Modifications to date here... https://github.com/bencoman/opensmalltalk-vm/tree/MinimalisticHeadless-x64-msvc2017 Next would be to catch up to Esteban's branch... https://github.com/estebanlm/opensmalltalk-vm/tree/add-minheadless-vm cheers -ben |
|
|
Hi Ben, I did not follow all the thread in detail, but a lot of missing definitions like alloca were in platform specific support headers: The fact that minheadless seems a
(better ?)
re-organisation of sources (I guess with a better separation of concerns) does not mean that we have to re-invent everything. It's just there under our hands. Le mar. 14 août 2018 à 07:46, Ben Coman <[hidden email]> a écrit :
|
|
|
On Wed, 5 Dec 2018 at 18:45, Nicolas Cellier <[hidden email]> wrote:
I don't think much is being reinvented. Its just minheadless branched before much work on 64-bit Windows. sqPlatformSpecific.h is very much reused, as borne out by this diff... .../platforms/win32/vm/sqPlatformSpecific.h .../platforms/minheadless/windows/sqPlatformSpecific-Win32.h /* win32 sqPlatformSpecific.h -- Platform-specific prototypes /* win32 sqPlatformSpecific.h -- Platform-specific prototypes /* How to use this file: /* How to use this file: This file is for general platform-specific macros and decl This file is for general platform-specific macros and decl The goal is to keep most of the other header files generic The goal is to keep most of the other header files generic To override a definition or macro from sq.h, you must firs To override a definition or macro from sq.h, you must firs provide the new definition. provide the new definition. */ */ #if _WIN32 || _WIN64 | #ifdef WIN32 /* Override necessary definitions */ /* Override necessary definitions */ #undef putchar #undef putchar #include "sqWin32Alloc.h" #include "sqWin32Alloc.h" #ifdef _MSC_VER #ifdef _MSC_VER #include <windows.h> #include <windows.h> #define HAVE_BOOLEAN 1 /* for jpegReaderWriter plugin compati #define HAVE_BOOLEAN 1 /* for jpegReaderWriter plugin compati #endif #endif #ifdef _MSC_VER #ifdef _MSC_VER #define squeakFileOffsetType __int64 #define squeakFileOffsetType __int64 #else #else #define squeakFileOffsetType unsigned long long #define squeakFileOffsetType unsigned long long #endif #endif #ifdef WIN32_FILE_SUPPORT #ifdef WIN32_FILE_SUPPORT #undef sqImageFile #undef sqImageFile #undef sqImageFileClose #undef sqImageFileClose #undef sqImageFileOpen #undef sqImageFileOpen #undef sqImageFilePosition #undef sqImageFilePosition #undef sqImageFileRead #undef sqImageFileRead #undef sqImageFileSeek #undef sqImageFileSeek #undef sqImageFileSeekEnd #undef sqImageFileSeekEnd #undef sqImageFileWrite #undef sqImageFileWrite #define sqImageFile usqIntptr_t #define sqImageFile usqIntptr_t sqInt sqImageFileClose(sqImageFile h); sqInt sqImageFileClose(sqImageFile h); sqImageFile sqImageFileOpen(const char *fileName, const char sqImageFile sqImageFileOpen(const char *fileName, const char squeakFileOffsetType sqImageFilePosition(sqImageFile h); squeakFileOffsetType sqImageFilePosition(sqImageFile h); size_t sqImageFileRead(void *ptr, size_t sz, size_t count, sq size_t sqImageFileRead(void *ptr, size_t sz, size_t count, sq squeakFileOffsetType sqImageFileSeek(sqImageFile h, squeakFil squeakFileOffsetType sqImageFileSeek(sqImageFile h, squeakFil squeakFileOffsetType sqImageFileSeekEnd(sqImageFile h, squeak squeakFileOffsetType sqImageFileSeekEnd(sqImageFile h, squeak size_t sqImageFileWrite(const void *ptr, size_t sz, size_t co size_t sqImageFileWrite(const void *ptr, size_t sz, size_t co #else /* when no WIN32_FILE_SUPPORT, add necessary stub for u #else /* when no WIN32_FILE_SUPPORT, add necessary stub for u #include <stdlib.h> #include <stdlib.h> #include <io.h> /* _get_osfhandle */ #include <io.h> /* _get_osfhandle */ > #ifndef PATH_MAX #define PATH_MAX _MAX_PATH #define PATH_MAX _MAX_PATH > #endif #define fsync(filenumber) FlushFileBuffers((HANDLE)_get_osfha #define fsync(filenumber) FlushFileBuffers((HANDLE)_get_osfha #endif /* WIN32_FILE_SUPPORT */ #endif /* WIN32_FILE_SUPPORT */ /* pluggable primitive support */ /* pluggable primitive support */ #if defined(_MSC_VER) || defined(__MINGW32__) #if defined(_MSC_VER) || defined(__MINGW32__) # undef EXPORT # undef EXPORT # define EXPORT(returnType) __declspec( dllexport ) returnTy # define EXPORT(returnType) __declspec( dllexport ) returnTy # undef VM_EXPORT # undef VM_EXPORT # define VM_EXPORT __declspec( dllexport ) # define VM_EXPORT __declspec( dllexport ) > # if defined(BUILD_VM_CORE) && !defined(VM_CORE_STATIC) > # undef VM_FUNCTION_EXPORT > # define VM_FUNCTION_EXPORT(returnType) __declspec( dllexport ) returnType > # else > # undef VM_FUNCTION_EXPORT > # define VM_FUNCTION_EXPORT(returnType) __declspec( dllimport ) returnType > # endif #endif #endif /* missing functions */ /* missing functions */ #ifdef _MSC_VER #ifdef _MSC_VER /* see on msdn the list of functions available /* see on msdn the list of functions available * CRT Alphabetical Function Reference * CRT Alphabetical Function Reference * https://msdn.microsoft.com/en-US/library/634ca0c2.aspx */ * https://msdn.microsoft.com/en-US/library/634ca0c2.aspx */ # include <malloc.h> # include <malloc.h> # include <float.h> # include <float.h> # ifndef alloca # ifndef alloca # define alloca _alloca # define alloca _alloca # endif # endif # if _MSC_VER < 1800 /* not available before MSVC 2013 */ # if _MSC_VER < 1800 /* not available before MSVC 2013 */ # define atoll(x) _atoi64(x) # define atoll(x) _atoi64(x) # define strtoll(beg,end,base) _strtoi64(beg,end,base) # define strtoll(beg,end,base) _strtoi64(beg,end,base) double round(double); double round(double); # endif # endif # if _MSC_VER < 1900 /* not available before MSVC 2015 */ # if _MSC_VER < 1900 /* not available before MSVC 2015 */ # define snprintf _snprintf # define snprintf _snprintf # ifndef isnan # ifndef isnan # define isnan _isnan # define isnan _isnan # endif # endif # endif # endif # if _MSC_VER < 1300 /* maybe not available before MSVC 7.0 # if _MSC_VER < 1300 /* maybe not available before MSVC 7.0 # define fabsf(x) ((float)fabs((double)(x))) # define fabsf(x) ((float)fabs((double)(x))) # endif # endif # define bzero(pointer,size) ZeroMemory(pointer,size) # define bzero(pointer,size) ZeroMemory(pointer,size) > #else > # include <string.h> > # define bzero(pointer,size) memset(pointer, 0, size) #endif #endif #ifdef __GNUC__ #ifdef __GNUC__ # if __GNUC__ < 3 # if __GNUC__ < 3 # define fabsf(x) ((float)fabs((double)(x))) /* not sur # define fabsf(x) ((float)fabs((double)(x))) /* not sur # endif # endif #endif #endif #else #else error "Not Win32 or Win64!" | error "Not Win32!" #endif /* _WIN32 || _WIN64 */ | #endif /* WIN32 */ int ioSetCursorARGB(sqInt bitsIndex, sqInt w, sqInt h, sqInt int ioSetCursorARGB(sqInt bitsIndex, sqInt w, sqInt h, sqInt /* poll and profile thread priorities. The stack vm uses a t /* poll and profile thread priorities. The stack vm uses a t * VM to poll for I/O, check for delay expiry et al at regula * VM to poll for I/O, check for delay expiry et al at regula * VMs use a thread to sample the pc for VM profiling. The p * VMs use a thread to sample the pc for VM profiling. The p * to have a priority higher than the main VM thread and the * to have a priority higher than the main VM thread and the * to have a priority higher than the poll thread to be able * to have a priority higher than the poll thread to be able * We would like POLL_THREAD_PRIORITY to be THREAD_PRIORITY_T * We would like POLL_THREAD_PRIORITY to be THREAD_PRIORITY_T * but SetThreadPriority fails with this value on Windows XP. * but SetThreadPriority fails with this value on Windows XP. * * * N.B. THREAD_PRIORITY_TIME_CRITICAL a.k.a. THREAD_BASE_PRIO * N.B. THREAD_PRIORITY_TIME_CRITICAL a.k.a. THREAD_BASE_PRIO * THREAD_PRIORITY_MAX a.k.a. THREAD_BASE_PRIORITY_MAX * THREAD_PRIORITY_MAX a.k.a. THREAD_BASE_PRIORITY_MAX * See WinBase.h & WinNT.h. * See WinBase.h & WinNT.h. */ */ #if STACKVM #if STACKVM # define POLL_THREAD_PRIORITY THREAD_PRIORITY_HIGHEST # define POLL_THREAD_PRIORITY THREAD_PRIORITY_HIGHEST #endif /* STACKVM */ #endif /* STACKVM */ #define PROF_THREAD_PRIORITY THREAD_PRIORITY_TIME_CRITICAL #define PROF_THREAD_PRIORITY THREAD_PRIORITY_TIME_CRITICAL #if COGVM #if COGVM extern void sqMakeMemoryExecutableFromTo(usqIntptr_t, usqIntp extern void sqMakeMemoryExecutableFromTo(usqIntptr_t, usqIntp extern void sqMakeMemoryNotExecutableFromTo(usqIntptr_t, usqI extern void sqMakeMemoryNotExecutableFromTo(usqIntptr_t, usqI extern int isCFramePointerInUse(void); extern int isCFramePointerInUse(void); extern int osCogStackPageHeadroom(void); extern int osCogStackPageHeadroom(void); extern void reportMinimumUnusedHeadroom(void); extern void reportMinimumUnusedHeadroom(void); #endif #endif /* Thread support for thread-safe signalSemaphoreWithIndex an /* Thread support for thread-safe signalSemaphoreWithIndex an #if STACKVM || NewspeakVM #if STACKVM || NewspeakVM # define sqLowLevelYield() Sleep(0) # define sqLowLevelYield() Sleep(0) /* these are used both in the STACKVM & the COGMTVM */ /* these are used both in the STACKVM & the COGMTVM */ # define sqOSThread void * # define sqOSThread void * # define ioOSThreadsEqual(a,b) ((a) == (b)) # define ioOSThreadsEqual(a,b) ((a) == (b)) # if COGMTVM # if COGMTVM /* Please read the comment for CogThreadManager in the VMMake /* Please read the comment for CogThreadManager in the VMMake * documentation of this API. * documentation of this API. */ */ # define sqOSSemaphore void * # define sqOSSemaphore void * # if !ForCOGMTVMImplementation /* this is a read-only export # if !ForCOGMTVMImplementation /* this is a read-only export extern const unsigned long tltiIndex; extern const unsigned long tltiIndex; # endif # endif # define ioGetThreadLocalThreadIndex() ((long)TlsGetValue(tl # define ioGetThreadLocalThreadIndex() ((long)TlsGetValue(tl # define ioSetThreadLocalThreadIndex(v) (TlsSetValue(tltiInd # define ioSetThreadLocalThreadIndex(v) (TlsSetValue(tltiInd # define ioTransferTimeslice() Sleep(0) # define ioTransferTimeslice() Sleep(0) # define ioMilliSleep(ms) Sleep(ms) # define ioMilliSleep(ms) Sleep(ms) # endif /* COGMTVM */ # endif /* COGMTVM */ #endif /* STACKVM || NewspeakVM */ #endif /* STACKVM || NewspeakVM */ #if defined(__GNUC__) #if defined(__GNUC__) # if !defined(VM_LABEL) # if !defined(VM_LABEL) # define VM_LABEL(foo) asm("\n.globl L" #foo "\nL" #foo # define VM_LABEL(foo) asm("\n.globl L" #foo "\nL" #foo # endif # endif #endif #endif #if !defined(VM_LABEL) || COGVM #if !defined(VM_LABEL) || COGVM # undef VM_LABEL # undef VM_LABEL # define VM_LABEL(foo) ((void)0) | # define VM_LABEL(foo) 0 #endif #endif /* Define the fields in a struct _CONTEXT as returned by GetThreadContext that * represent the program counter and frame pointer on the current architecture. */ #if defined(_M_IX86) || defined(_M_I386) || defined(_X86_) || defined(i386) || defined(__i386__) # define CONTEXT_PC Eip # define CONTEXT_FP Ebp # define CONTEXT_SP Esp #elif defined(x86_64) || defined(__x86_64) || defined(__x86_64__) || defined(__amd64) || defined(__amd64__) || defined(x64) || defined(_M_AMD64) || defined(_M_X64) || defined(_M_IA64) # define CONTEXT_PC Rip # define CONTEXT_FP Rbp # define CONTEXT_SP Rsp #else # error "unknown architecture, program counter field undefined" #endif > #ifdef _MSC_VER > /* disable "function XXXX: no return value" */ > #pragma warning(disable:4035) > /* optional C SEH macros */ > #define TRY __try > #define EXCEPT(filter) __except(filter) > #define FINALLY __finally > #else > /* optional C SEH macros */ > #define TRY > #define EXCEPT(filter) if (0) > #define FINALLY > #endif But perhaps instead of my change to "platforms/Cross/plugins/IA32ABI/x64win64abicc.c" here... it would have been better to include "sqPlatformSpecific.h" ? Or would that be heavy handed? I don't know. At the time it seemed reasonable to echo the similar define here... platforms/Cross/plugins/IA32ABI/ia32abicc.c: # define alloca _alloca cheers -ben P.S. It might be good to reconcile the two files, but I don't know enough about all the changes myself. And to avoid divergence maybe(??) one file should #include the other instead of being a copy. Or would that be a bad?? |
«
Return to Squeak VM
|
1 view|%1 views
| Free forum by Nabble | Edit this page |