Remove MIUtilParse (no longer used)

[mgorny]

Summary

Remove MIUtilParse once https://reviews.llvm.org/D23882 is accepted.

[mgorny]

@labath, since you committed D23882, could you, please, take a look at this one as well? It's immediately following that, and should fix the build issue resulting from use of private headers.

[krytarowski]

It looks good. Remaining TODO for standalone builds:

• sanitize six.py usage (it's installed into system-wide directory, where standard py-six lands)
• bump cmake_minimum_required(VERSION 2.8.12.2) to 3.4.3 in cmake/modules/LLDBStandalone.cmake

Final bits for Windows:

• eliminate the usage of ../lib/Support/regex_impl.h in RegularExpression.h

[mgorny]

In D23883#539442, @krytarowski wrote:

It looks good. Remaining TODO for standalone builds:

• sanitize six.py usage (it's installed into system-wide directory, where standard py-six lands)

How would you prefer handling it? Can we just kill it and rely on user installing it system-wide or via pip?

• bump cmake_minimum_required(VERSION 2.8.12.2) to 3.4.3 in cmake/modules/LLDBStandalone.cmake

Final bits for Windows:

• eliminate the usage of ../lib/Support/regex_impl.h in RegularExpression.h

Wouldn't it be actually better to kill that regex implementation as well, and use the class provided by LLVMSupport everywhere? I started with the other since it caused direct build issues on Linux but I think I could try to kill this one as well.

[krytarowski]

In D23883#539471, @mgorny wrote:

In D23883#539442, @krytarowski wrote:

It looks good. Remaining TODO for standalone builds:

• sanitize six.py usage (it's installed into system-wide directory, where standard py-six lands)

How would you prefer handling it? Can we just kill it and rely on user installing it system-wide or via pip?

1. Patch scripts/Python/finishSwigPythonLLDB.py and install six into lldb subdir with a custom name (lldb_six.py).
2. Patch the Python sources to make use of lldb_six.

Unfortunately there is resistance to grab system-wide (provided by a user) six module.

• bump cmake_minimum_required(VERSION 2.8.12.2) to 3.4.3 in cmake/modules/LLDBStandalone.cmake

Final bits for Windows:

• eliminate the usage of ../lib/Support/regex_impl.h in RegularExpression.h

Wouldn't it be actually better to kill that regex implementation as well, and use the class provided by LLVMSupport everywhere? I started with the other since it caused direct build issues on Linux but I think I could try to kill this one as well.

It makes sense.

[ki.stfu]

You forgot to remove its header file

[mgorny]

In D23883#539611, @ki.stfu wrote:

You forgot to remove its header file

Thanks for noticing. Fixed now.

[labath]

@mgorny: The change should be reviewed by an lldb-mi maintainer (i.e., @ki.stfu). It helps is you explicitly specify the reviewer, as otherwise the person may not notice the patch.

In D23883#539471, @mgorny wrote:

In D23883#539442, @krytarowski wrote:

It looks good. Remaining TODO for standalone builds:

• sanitize six.py usage (it's installed into system-wide directory, where standard py-six lands)

How would you prefer handling it? Can we just kill it and rely on user installing it system-wide or via pip?

I think the most politically passable version would be to make that a cmake option (LLDB_USE_BUILTIN_SIX ?). If it is set you use the system-wide package, otherwise, you use the builtin one. When you build the distro package, you set it to false. When e.g., we are distributing it with Android Studio, we'll set it to true. I personally don't care what the default will be.

• bump cmake_minimum_required(VERSION 2.8.12.2) to 3.4.3 in cmake/modules/LLDBStandalone.cmake

Final bits for Windows:

• eliminate the usage of ../lib/Support/regex_impl.h in RegularExpression.h

Wouldn't it be actually better to kill that regex implementation as well, and use the class provided by LLVMSupport everywhere? I started with the other since it caused direct build issues on Linux but I think I could try to kill this one as well.

I think that would be great. We'll need to be a bit careful and check whether we the two libraries have sufficiently similar interfaces.

[ki.stfu]

lgtm

[labath]

I am going to assume you still want me to submit this. r281317.

[mgorny]

Yes, thanks a lot.

tools/lldb-mi/CMakeLists.txt

set(LLDB_MI_SOURCES
   MICmnStreamStderr.cpp
   MICmnStreamStdin.cpp
   MICmnStreamStdout.cpp
   MICmnThreadMgrStd.cpp
   MIDriver.cpp
   MIDriverBase.cpp
   MIDriverMain.cpp
   MIDriverMgr.cpp
-  MIUtilParse.cpp
   MIUtilDateTimeStd.cpp
   MIUtilDebug.cpp
   MIUtilFileStd.cpp
   MIUtilMapIdToVariant.cpp
   MIUtilString.cpp
   MIUtilThreadBaseStd.cpp
   MIUtilVariant.cpp
   )

tools/lldb-mi/MIUtilParse.cpp

-//===-- MIUtilParse.cpp ----------------------------------------*- C++ -*-===//
-//
-//                     The LLVM Compiler Infrastructure
-//
-// This file is distributed under the University of Illinois Open Source
-// License. See LICENSE.TXT for details.
-//
-//===----------------------------------------------------------------------===//
- 
-// Third party headers:
-#include <memory>
-
-// In-house headers:
-#include "MIUtilParse.h"
- 
-//++ ------------------------------------------------------------------------------------
-// Details: CRegexParser constructor.
-// Type:    Method.
-// Args:    regexStr - Pointer to the regular expression to compile.
-// Return:  None.
-// Throws:  None.
-//--
-MIUtilParse::CRegexParser::CRegexParser(const char *regexStr)
-    : m_isValid(llvm_regcomp(&m_emma, regexStr, REG_EXTENDED) == 0)
-{
-}
- 
-//++ ------------------------------------------------------------------------------------
-// Details: CRegexParser destructor.
-// Type:    Method.
-// Args:    None.
-// Return:  None.
-// Throws:  None.
-//--
-MIUtilParse::CRegexParser::~CRegexParser()
-{
-    // Free up memory held within regex.
-    if (m_isValid)
-        llvm_regfree(&m_emma);
-}
- 
-//++ ------------------------------------------------------------------------------------
-// Details: CRegexParser regex executer.
-//          Match the input against the regular expression.  Return an error
-//          if the number of matches is less than minMatches.  If the default
-//          minMatches value of 0 is passed, an error will be returned if
-//          the number of matches is less than the maxMatches value used to
-//          initialize Match.
-// Type:    Method.
-// Args:    input (R) - Pointer to UTF8 text data to be parsed.
-//          match (RW) - Reference to Match class.
-//          minMatches (R) - Minimum number of regex matches expected.
-// Return:  bool - True = minimum matches were met,
-//                 false = minimum matches were not met or regex failed.
-// Throws:  None.
-//--
-bool
-MIUtilParse::CRegexParser::Execute(const char *input, Match& match, size_t minMatches)
-{
-    if (!m_isValid)
-        return false;
- 
-    std::unique_ptr<llvm_regmatch_t[]> matches(new llvm_regmatch_t[match.m_maxMatches]); // Array of matches
-   
-    if (llvm_regexec(&m_emma, input, match.m_maxMatches, matches.get(), 0) != 0)
-        return false;
- 
-    size_t i;
-    for (i = 0; i < match.m_maxMatches && matches[i].rm_so >= 0; i++)
-    {
-        const int n = matches[i].rm_eo - matches[i].rm_so;
-        match.m_matchStrs[i].assign(input + matches[i].rm_so, n);
-    }
-    return i >= minMatches;
-}