Various build fixes for lldb on MinGW

[mstorsjo]

As context (I haven't followed lldb almost at all), what's the general status of lldb on windows - generally working, or still a work in progress? Does it work with dwarf debug info in COFF (as commonly used in mingw environments), in addition to PDB? I.e. is the existing dwarf support tied to the ELF/MachO formats, or is it decoupled?

[clayborg]

In D65691#1613395, @mstorsjo wrote:

As context (I haven't followed lldb almost at all), what's the general status of lldb on windows - generally working, or still a work in progress? Does it work with dwarf debug info in COFF (as commonly used in mingw environments), in addition to PDB? I.e. is the existing dwarf support tied to the ELF/MachO formats, or is it decoupled?

It is decoupled. As long as the ObjectFileCOFF hands out the right section for the section enumeration in lldb-enumerations.h in the lldb::SectionType, it should all just work. The enumerations in question are all of the enums that start with eSectionTypeDWARF. It seems to have support in ObjectFilePECOFF::CreateSections(...). So I think it should all just work. Let us know if there are any issues.

[xbolva00]

lldb/source/Plugins/Process/gdb-remote/ProcessGDBRemote.cpp

19 ↗

(On Diff #213175)

Maybe drop usleep and use C++’s

https://en.cppreference.com/w/cpp/thread/sleep_for ?

[labath]

I took @xbolva00's suggestion one step further, and removed the usleep and the related compat code in r367814. So, some of your changes are no longer needed once you rebase past that.

I have a couple of additional small comments inline, but other than that, this seems fine. I'm not sure how active is @zturner these days (@amccarth and @stella.stamenova are probably better reviewers for windows stuff now), but this seems pretty straight-forward, so I don't think we have to wait for them.

lldb/source/Initialization/SystemInitializerCommon.cpp

28

In r346625, we deleted all/most comments like this. Probably best to not introduce new ones...

lldb/tools/driver/Platform.h

20

It looks like we already have the HAVE_SYS_TYPES_H macro, which we can use for this purpose, so I'd use that, and move it out of the _MSC_VER block as we should really include this header on other platforms too (but we probably haven't done that because it got included transitively already...)

[amccarth]

Yes, zturner isn't as active here as he was before.

I'm happy to review patches regarding LLDB on Windows, even though I don't know much about Mingw.

lldb/source/Plugins/Process/gdb-remote/ProcessGDBRemote.cpp

19 ↗

(On Diff #213175)

+1. usleep is deprecated in POSIX and is essentially meaningless on Windows (in terms of how the scheduler works) unless you're waiting 1000 or more microseconds. At this point, sleep_for is likely to be better supported by the platforms.

lldb/tools/lldb-vscode/VSCode.cpp

20

I would suggest moving the #ifndef __MINGW32__ up a line to keep the NOMINMAX adjacent to the include of Windows.h. NOMINMAX is relevant only to Windows.h.

lldb/tools/lldb-vscode/lldb-vscode.cpp

55

Nothing in the rest of this .cpp file uses PATH_MAX, so just delete the #define instead of executing it conditionally.

[hhb]

Fix comments.

lldb/tools/lldb-vscode/lldb-vscode.cpp

55

PATH_MAX is used in SendProcessEvent (line 283)

[amccarth]

lldb/tools/lldb-vscode/lldb-vscode.cpp

55

Got it. When I asked Phabricator to show me all the lines, it didn't show me _all_ the lines.

That said, it seems unfortunate that it's yet another fixed-length filename buffer instead of getting a ConstString from the FileSpec. But that's outside the scope of this change.

[hhb]

Rebase

[labath]

lldb/tools/lldb-vscode/lldb-vscode.cpp

55

The problem is that you're dealing with the SBFileSpec here, and that makes things slightly complicated (stable api, not being able to return c++ objects through it, etc...).

[jingham]

lldb/tools/lldb-vscode/lldb-vscode.cpp

55

There's no reason we couldn't add an SBString class to vend non-fixed length strings. There haven't been enough instances where we really needed it to motivate that, but we have SBStringList to wrap vectors of strings, so it wouldn't be out of place.

lldb/include/lldb/Host/windows/PosixApi.h

 #define S_IXGRP 0 /* execute, group */
 #define S_IROTH 0 /* read, others */
 #define S_IWOTH 0 /* write, others */
 #define S_IXOTH 0 /* execute, others */
 #define S_IRWXU 0
 #define S_IRWXG 0
 #define S_IRWXO 0
 
+#ifdef __MINGW32__
+// pyconfig.h typedefs this.  We require python headers to be included before
+// any LLDB headers, but there's no way to prevent python's pid_t definition
+// from leaking, so this is the best option.
+#ifndef NO_PID_T
+#include <sys/types.h>
+#endif
+#endif // __MINGW32__
+
 #ifdef _MSC_VER
 
 // PRIxxx format macros for printf()
 #include <inttypes.h>
 
 // open(), close(), creat(), etc.
 #include <io.h>
 

 #endif // _MSC_VER
 
 // Various useful posix functions that are not present in Windows.  We provide
 // custom implementations.
 int vasprintf(char **ret, const char *fmt, va_list ap);
 char *strcasestr(const char *s, const char *find);
 char *realpath(const char *name, char *resolved);
 
+#ifdef _MSC_VER
+
 int usleep(uint32_t useconds);
 char *basename(char *path);
 char *dirname(char *path);
 
 int strcasecmp(const char *s1, const char *s2);
 int strncasecmp(const char *s1, const char *s2, size_t n);
 
+#endif // _MSC_VER
+
 // empty functions
 inline int posix_openpt(int flag) { LLVM_BUILTIN_UNREACHABLE; }
 
 inline int strerror_r(int errnum, char *buf, size_t buflen) {
   LLVM_BUILTIN_UNREACHABLE;
 }
 
 inline int unlockpt(int fd) { LLVM_BUILTIN_UNREACHABLE; }
 inline int grantpt(int fd) { LLVM_BUILTIN_UNREACHABLE; }
 inline char *ptsname(int fd) { LLVM_BUILTIN_UNREACHABLE; }
 
 inline pid_t fork(void) { LLVM_BUILTIN_UNREACHABLE; }
 inline pid_t setsid(void) { LLVM_BUILTIN_UNREACHABLE; }
 
 #endif

lldb/source/Host/windows/FileSystem.cpp

 //===-- FileSystem.cpp ------------------------------------------*- C++ -*-===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 
 #include "lldb/Host/windows/windows.h"
 
+#include <share.h>
 #include <shellapi.h>
 #include <sys/stat.h>
 #include <sys/types.h>
 
 #include "lldb/Host/FileSystem.h"
 #include "lldb/Host/windows/AutoHandle.h"
 #include "lldb/Host/windows/PosixApi.h"
 

lldb/source/Initialization/SystemInitializerCommon.cpp

 
 #if defined(__linux__) || defined(__FreeBSD__) || defined(__NetBSD__)
 #include "Plugins/Process/POSIX/ProcessPOSIXLog.h"
 #endif
 
 #if defined(_WIN32)
 #include "Plugins/Process/Windows/Common/ProcessWindowsLog.h"
 #include "lldb/Host/windows/windows.h"
+#include <crtdbg.h>
 #endif
 
 #include "llvm/Support/TargetSelect.h"
 
 #include <string>
 
 using namespace lldb_private;
 using namespace lldb_private::repro;

lldb/tools/driver/Platform.h

 #define lldb_Platform_h_
 
 #if defined(_WIN32)
 
 #include <io.h>
 #if defined(_MSC_VER)
 #include <signal.h>
 #endif
+#if HAVE_SYS_TYPES_H
+#include <sys/types.h>
+#endif
 #include "lldb/Host/windows/windows.h"
 #include <inttypes.h>
 
 struct winsize {
   long ws_col;
 };
 
 typedef unsigned char cc_t;

lldb/tools/lldb-vscode/VSCode.cpp

 #include <fstream>
 #include <mutex>
 
 #include "LLDBUtils.h"
 #include "VSCode.h"
 #include "llvm/Support/FormatVariadic.h"
 
 #if defined(_WIN32)
+#ifndef __MINGW32__
 #define NOMINMAX
 #include <Windows.h>
+#endif // __MINGW32__
 #include <fcntl.h>
 #include <io.h>
 #endif
 
 using namespace lldb_vscode;
 
 namespace lldb_vscode {
 

lldb/tools/lldb-vscode/lldb-vscode.cpp

 #include "llvm/Support/FileSystem.h"
 #include "llvm/Support/raw_ostream.h"
 
 #include "JSONUtils.h"
 #include "LLDBUtils.h"
 #include "VSCode.h"
 
 #if defined(_WIN32)
+#ifndef PATH_MAX
 #define PATH_MAX MAX_PATH
+#endif
 typedef int socklen_t;
 constexpr const char *dev_null_path = "nul";
 
 #else
 constexpr const char *dev_null_path = "/dev/null";
 
 #endif