From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from us-smtp-delivery-124.mimecast.com (us-smtp-delivery-124.mimecast.com [170.10.133.124]) by sourceware.org (Postfix) with ESMTPS id 0A24E3858D28 for ; Thu, 25 Nov 2021 10:36:52 +0000 (GMT) DMARC-Filter: OpenDMARC Filter v1.4.1 sourceware.org 0A24E3858D28 Received: from mail-wm1-f70.google.com (mail-wm1-f70.google.com [209.85.128.70]) by relay.mimecast.com with ESMTP with STARTTLS (version=TLSv1.2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id us-mta-264-guA2ru8oOmy0rQJWw-EevA-1; Thu, 25 Nov 2021 05:36:50 -0500 X-MC-Unique: guA2ru8oOmy0rQJWw-EevA-1 Received: by mail-wm1-f70.google.com with SMTP id r129-20020a1c4487000000b00333629ed22dso4720714wma.6 for ; Thu, 25 Nov 2021 02:36:50 -0800 (PST) X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:from:to:cc:subject:date:message-id:in-reply-to :references:mime-version:content-transfer-encoding; bh=znR6tmMkzP9P11Oe3SFxQBRmERo/rjP9AH2cJAM9q6s=; b=LZh2K6GpJ/D9CiCdHDwhHKrHZj/0HhE7DK+7jUl2JekLj2Vtjs2s3CfboB/ZVVvrtA nastKY5UaLMoclzWibEJrcwYoFnXHGQysK2rhhe7qkwycQhVqlroi1awn9t+bCr/9F6X w769dmOclFDhkeVbToK1t6/XWctEkt3OktmyzDoYM1LCg4im5BgEzAnjknbwUp99xv8R MVQuCP0mTu54jIn0cOivX+pV4jI4QfyO4Ico7xW/tayNp8d+X19yJ70JpNQoA1+g9omT +v+0SifBaVOLaGK8QUZX+2CUw1qx0jdI13+DYIGnKasWwXoAUaGDnBs7i2CVTw4WuG/7 KfVw== X-Gm-Message-State: AOAM531Cl6dWSRWQXRLsBxB5+gJSkqFxkpkfj+GRYQ0Vv5YPDSzvuaGs MPJiI8iaEbqIIGUqWUGk7xUMT0TL9uizlayt32bn4dTcrDc5mLEyIVcNhJo/V6v4VZ2ZbVCK09w Qs+LC7i5T/iOzoqsTFQR1Mq4fbcJ8Qs3GHd6puXreBhXkpbULTJtbtJkPQTQNmL4KZhWUf9A5OA == X-Received: by 2002:a1c:4b07:: with SMTP id y7mr5681242wma.188.1637836608376; Thu, 25 Nov 2021 02:36:48 -0800 (PST) X-Google-Smtp-Source: ABdhPJxHzU/SzR5YiJnY7EVczQf1HGfcjOZhrNIU5b63JhZWiChiVLPBRIreyspZvLBUwNhlUjhRLA== X-Received: by 2002:a1c:4b07:: with SMTP id y7mr5681159wma.188.1637836607675; Thu, 25 Nov 2021 02:36:47 -0800 (PST) Received: from localhost (host86-166-129-255.range86-166.btcentralplus.com. [86.166.129.255]) by smtp.gmail.com with ESMTPSA id g13sm3526175wrd.57.2021.11.25.02.36.47 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Thu, 25 Nov 2021 02:36:47 -0800 (PST) From: Andrew Burgess To: gdb-patches@sourceware.org Cc: Andrew Burgess Subject: [PATCHv2 1/2] gdb: use python to colorize disassembler output Date: Thu, 25 Nov 2021 10:36:42 +0000 Message-Id: <9cd73a16d61b3ddcc277093b4097f9c057e64922.1637836489.git.aburgess@redhat.com> X-Mailer: git-send-email 2.25.4 In-Reply-To: References: MIME-Version: 1.0 X-Mimecast-Spam-Score: 0 X-Mimecast-Originator: redhat.com Content-Transfer-Encoding: 8bit Content-Type: text/plain; charset="US-ASCII" X-Spam-Status: No, score=-9.2 required=5.0 tests=BAYES_00, DKIM_INVALID, DKIM_SIGNED, GIT_PATCH_0, KAM_DMARC_NONE, KAM_DMARC_STATUS, RCVD_IN_BARRACUDACENTRAL, RCVD_IN_DNSWL_LOW, RCVD_IN_MSPIKE_H2, SPF_HELO_NONE, SPF_NONE, TXREP autolearn=ham autolearn_force=no version=3.4.4 X-Spam-Checker-Version: SpamAssassin 3.4.4 (2020-01-24) on server2.sourceware.org X-BeenThere: gdb-patches@sourceware.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: Gdb-patches mailing list List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Thu, 25 Nov 2021 10:36:54 -0000 From: Andrew Burgess This commit adds styling support to the disassembler output, as such two new commands are added to GDB: set style disassembler on|off show style disassembler In this commit I make use of the Python Pygments package to provide the styling. I did investigate making use of libsource-highlight, however, I found the highlighting results to be inferior to those of Pygments; only some mnemonics were highlighted, and highlighting of register names such as r9d and r8d (on x86-64) was incorrect. To enable disassembler highlighting via Pygments, I've added a new extension language hook, which is then implemented for Python. This hook is very similar to the existing hook for source code colorization. One possibly odd choice I made with the new hook is to pass a gdb.Architecture through, even though this is currently unused. The reason this argument is not used is that, currently, styling is performed identically for all architectures. However, even though the Python function used to perform styling of disassembly output is not part of any undocumented API, I don't want to close the door on a user overriding this function to provide architecture specific styling. To do this, the user would inevitably require access to the gdb.Architecture, and so I decided to add this field now. The styling is applied within gdb_disassembler::print_insn, to achieve this, gdb_disassembler now writes its output into a temporary buffer, styling is then applied to the contents of this buffer. Finally the gdb_disassembler buffer is copied out to its final destination stream. There's a new test to check that the disassembler output includes some escape sequences, though I don't check for specific colours; the precise colors will depend on which instructions are in the disassembler output. The only negative change with this commit relates to how addresses are printed, in the case when the Python Pygments package is not available. Addresses are printed via calls to GDB's print_address function. Traditionally, this would provide styling for the address and symbol name if the output ui_file* supported styling. Now that we want to apply styling after the disassembler has finished, all disassembler output is written into a temporary string_file, which is configured not to support styling. As a result, the print_address call no longer performs styling. If Pygments is available, this isn't a huge problem, the output will be fully styled one the disassembler has finished. However, if Pygments is not available, or fails for some reason, then, it is now too late to go back and have print_address apply styling. We have lost all print_address styling in this case. I don't know how much of a problem this is, for me, having the disassembler fully styled is a big enough win. But, if people see this as a huge problem we can investigate mechanisms to restore the print_address styling (for the case where Pygments is not available). --- gdb/NEWS | 6 +++ gdb/cli/cli-style.c | 26 +++++++++++++ gdb/cli/cli-style.h | 3 ++ gdb/disasm.c | 59 +++++++++++++++++++++++++++- gdb/disasm.h | 25 ++++++++++++ gdb/doc/gdb.texinfo | 15 +++++++ gdb/extension-priv.h | 6 +++ gdb/extension.c | 20 ++++++++++ gdb/extension.h | 8 ++++ gdb/python/lib/gdb/__init__.py | 11 ++++++ gdb/python/python.c | 67 ++++++++++++++++++++++++++++++++ gdb/testsuite/gdb.base/style.exp | 44 ++++++++++++++++++--- 12 files changed, 282 insertions(+), 8 deletions(-) diff --git a/gdb/NEWS b/gdb/NEWS index b0e4cce72ee..c62fed3f37f 100644 --- a/gdb/NEWS +++ b/gdb/NEWS @@ -39,6 +39,12 @@ set logging enabled on|off show logging enabled These commands set or show whether logging is enabled or disabled. +set style disassembly on|off +show style disassembly + If GDB is compiled with Python support, and the Python Pygments + package is available, then, when this setting is on, disassembler + output will have styling applied. + * Python API ** New function gdb.add_history(), which takes a gdb.Value object diff --git a/gdb/cli/cli-style.c b/gdb/cli/cli-style.c index a94b8c940bc..26abc3374ca 100644 --- a/gdb/cli/cli-style.c +++ b/gdb/cli/cli-style.c @@ -38,6 +38,11 @@ bool cli_styling = true; bool source_styling = true; +/* True if disassembler styling is enabled. Note that this is only + consulted when cli_styling is true. */ + +bool disassembler_styling = true; + /* Name of colors; must correspond to ui_file_style::basic_color. */ static const char * const cli_colors[] = { "none", @@ -301,6 +306,18 @@ show_style_sources (struct ui_file *file, int from_tty, fprintf_filtered (file, _("Source code styling is disabled.\n")); } +/* Implement 'show style disassembler'. */ + +static void +show_style_disassembler (struct ui_file *file, int from_tty, + struct cmd_list_element *c, const char *value) +{ + if (disassembler_styling) + fprintf_filtered (file, _("Disassembler output styling is enabled.\n")); + else + fprintf_filtered (file, _("Disassembler output styling is disabled.\n")); +} + void _initialize_cli_style (); void _initialize_cli_style () @@ -337,6 +354,15 @@ available if the appropriate extension is available at runtime." ), set_style_enabled, show_style_sources, &style_set_list, &style_show_list); + add_setshow_boolean_cmd ("disassembler", no_class, &disassembler_styling, _("\ +Set whether disassembler output styling is enabled."), _("\ +Show whether disassembler output styling is enabled."), _("\ +If enabled, disassembler output is styled. Disassembler highlighting\n\ +requires the Python Pygments library, if this library is not available\n\ +then disassembler highlighting will not be possible." + ), set_style_enabled, show_style_disassembler, + &style_set_list, &style_show_list); + file_name_style.add_setshow_commands (no_class, _("\ Filename display styling.\n\ Configure filename colors and display intensity."), diff --git a/gdb/cli/cli-style.h b/gdb/cli/cli-style.h index 78bc2cd6f1e..5361a644a95 100644 --- a/gdb/cli/cli-style.h +++ b/gdb/cli/cli-style.h @@ -128,6 +128,9 @@ extern cli_style_option version_style; /* True if source styling is enabled. */ extern bool source_styling; +/* True if disassembler styling is enabled. */ +extern bool disassembler_styling; + /* True if styling is enabled. */ extern bool cli_styling; diff --git a/gdb/disasm.c b/gdb/disasm.c index c045dfc94a6..8f7ec00486e 100644 --- a/gdb/disasm.c +++ b/gdb/disasm.c @@ -753,9 +753,12 @@ get_all_disassembler_options (struct gdbarch *gdbarch) gdb_disassembler::gdb_disassembler (struct gdbarch *gdbarch, struct ui_file *file, di_read_memory_ftype read_memory_func) - : m_gdbarch (gdbarch) + : m_gdbarch (gdbarch), + m_buffer (!use_ext_lang_colorization_p && disassembler_styling + && file->can_emit_style_escape ()), + m_dest (file) { - init_disassemble_info (&m_di, file, dis_asm_fprintf); + init_disassemble_info (&m_di, &m_buffer, dis_asm_fprintf); m_di.flavour = bfd_target_unknown_flavour; m_di.memory_error_func = dis_asm_memory_error; m_di.print_address_func = dis_asm_print_address; @@ -784,14 +787,66 @@ gdb_disassembler::~gdb_disassembler () disassemble_free_target (&m_di); } +/* See disasm.h. */ + +bool gdb_disassembler::use_ext_lang_colorization_p = true; + +/* See disasm.h. */ + int gdb_disassembler::print_insn (CORE_ADDR memaddr, int *branch_delay_insns) { m_err_memaddr.reset (); + m_buffer.clear (); int length = gdbarch_print_insn (arch (), memaddr, &m_di); + /* If we have successfully disassembled an instruction, styling is + on, we think that the extension language might be able to perform + styling for us, and the destination can support styling, then lets + call into the extension languages in order to style this output. */ + if (length > 0 && disassembler_styling + && use_ext_lang_colorization_p + && m_dest->can_emit_style_escape ()) + { + gdb::optional ext_contents; + ext_contents = ext_lang_colorize_disasm (m_buffer.string (), arch ()); + if (ext_contents.has_value ()) + m_buffer.string () = std::move (*ext_contents); + else + { + /* The extension language failed to add styling to the + disassembly output. Set the static flag so that next time we + disassemble we don't even bother attempting to use the + extension language for styling. */ + use_ext_lang_colorization_p = false; + + /* The instruction we just disassembled, and the extension + languages failed to highlight, might have otherwise had some + minimal highlighting applied by GDB. In regain that + highlighting we need to recreate m_buffer, but this time with + styling support. + + To do this we perform an in-place new, but this time turn on + the styling support, then we can re-disassembly the + instruction, and gain any minimal styling GDB might add. */ + gdb_static_assert ((std::is_same::value)); + gdb_assert (!m_buffer.term_out ()); + m_buffer.~string_file (); + new (&m_buffer) string_file (true); + length = gdbarch_print_insn (arch (), memaddr, &m_di); + gdb_assert (length > 0); + } + } + + /* Push any disassemble output to the real destination stream. We do + this even if the disassembler reported failure (-1) as the + disassembler may have printed something to its output stream. */ + m_di.fprintf_func (m_dest, "%s", m_buffer.c_str ()); + + /* If the disassembler failed then report an appropriate error. */ if (length < 0) { if (m_err_memaddr.has_value ()) diff --git a/gdb/disasm.h b/gdb/disasm.h index f6de33e3db8..cb7f3768995 100644 --- a/gdb/disasm.h +++ b/gdb/disasm.h @@ -82,6 +82,31 @@ class gdb_disassembler non-memory error. */ gdb::optional m_err_memaddr; + /* Disassembler output is built up into this buffer. Whether this + string_file is created with styling support or not depends on the + value of use_ext_lang_colorization_p, as well as whether disassembler + styling in general is turned on, and also, whether *m_dest supports + styling or not. */ + string_file m_buffer; + + /* The stream to which disassembler output will be written. */ + ui_file *m_dest; + + /* When true, m_buffer will be created without styling support, + otherwise, m_buffer will be created with styling support. + + This field will initially be true, but will be set to false if + ext_lang_colorize_disasm fails to add styling at any time. + + If the extension language is going to add the styling then m_buffer + should be created without styling support, the extension language will + then add styling at the end of the disassembly process. + + If the extension language is not going to add the styling, then we + create m_buffer with styling support, and GDB will add minimal styling + (currently just to addresses and symbols) as it goes. */ + static bool use_ext_lang_colorization_p; + static int dis_asm_fprintf (void *stream, const char *format, ...) ATTRIBUTE_PRINTF(2,3); diff --git a/gdb/doc/gdb.texinfo b/gdb/doc/gdb.texinfo index 1a944b10ae5..458822577d7 100644 --- a/gdb/doc/gdb.texinfo +++ b/gdb/doc/gdb.texinfo @@ -26080,6 +26080,21 @@ @item show style sources Show the current state of source code styling. + +@item set style disassembly @samp{on|off} +Enable or disable disassembly styling. This affects whether +disassembly output, such as the output of the @code{disassemble} +command, is styled. Disassembly styling only works if styling in +general is enabled (with @code{set style enabled on}), and if a source +highlighting library is available to @value{GDBN}. + +To highlight disassembler output, @value{GDBN} must be compiled with +Python support, and the Python Pygments package must be available. If +these requirements are not met then @value{GDBN} will not highlight +disassembler output, even when this option is @samp{on}. + +@item show style disassembly +Show the current state of disassembly styling. @end table Subcommands of @code{set style} control specific forms of styling. diff --git a/gdb/extension-priv.h b/gdb/extension-priv.h index 77f23e0f911..b2150624dde 100644 --- a/gdb/extension-priv.h +++ b/gdb/extension-priv.h @@ -257,6 +257,12 @@ struct extension_language_ops or an empty option. */ gdb::optional (*colorize) (const std::string &name, const std::string &contents); + + /* Colorize a single line of disassembler output, CONTENT. This should + either return colorized (using ANSI terminal escapes) version of the + contents, or an empty optional. */ + gdb::optional (*colorize_disasm) (const std::string &content, + gdbarch *gdbarch); }; /* State necessary to restore a signal handler to its previous value. */ diff --git a/gdb/extension.c b/gdb/extension.c index 89ab29f3d1c..6862147ac2f 100644 --- a/gdb/extension.c +++ b/gdb/extension.c @@ -904,6 +904,26 @@ ext_lang_colorize (const std::string &filename, const std::string &contents) return result; } +/* See extension.h. */ + +gdb::optional +ext_lang_colorize_disasm (const std::string &content, gdbarch *gdbarch) +{ + gdb::optional result; + + for (const struct extension_language_defn *extlang : extension_languages) + { + if (extlang->ops == nullptr + || extlang->ops->colorize_disasm == nullptr) + continue; + result = extlang->ops->colorize_disasm (content, gdbarch); + if (result.has_value ()) + return result; + } + + return result; +} + /* Called via an observer before gdb prints its prompt. Iterate over the extension languages giving them a chance to change the prompt. The first one to change the prompt wins, diff --git a/gdb/extension.h b/gdb/extension.h index 2f2ca3e7743..e8d2fbc7fc3 100644 --- a/gdb/extension.h +++ b/gdb/extension.h @@ -319,6 +319,14 @@ extern void get_matching_xmethod_workers extern gdb::optional ext_lang_colorize (const std::string &filename, const std::string &contents); +/* Try to colorize a single line of disassembler output, CONTENT for + GDBARCH. This will return either a colorized (using ANSI terminal + escapes) version of CONTENT, or an empty value if colorizing could not + be done. */ + +extern gdb::optional ext_lang_colorize_disasm + (const std::string &content, gdbarch *gdbarch); + #if GDB_SELF_TEST namespace selftests { extern void (*hook_set_active_ext_lang) (); diff --git a/gdb/python/lib/gdb/__init__.py b/gdb/python/lib/gdb/__init__.py index 7b6d8701548..6ab797c8c5e 100644 --- a/gdb/python/lib/gdb/__init__.py +++ b/gdb/python/lib/gdb/__init__.py @@ -243,8 +243,19 @@ try: except: return None + def colorize_disasm(content, gdbarch): + # Don't want any errors. + try: + lexer = lexers.get_lexer_by_name("asm") + formatter = formatters.TerminalFormatter() + return highlight(content, lexer, formatter).rstrip() + except: + return None except: def colorize(filename, contents): return None + + def colorize_disasm(content, gdbarch): + return None diff --git a/gdb/python/python.c b/gdb/python/python.c index d8a6a5978de..92619fdb3ad 100644 --- a/gdb/python/python.c +++ b/gdb/python/python.c @@ -121,6 +121,8 @@ static enum ext_lang_rc gdbpy_before_prompt_hook (const struct extension_language_defn *, const char *current_gdb_prompt); static gdb::optional gdbpy_colorize (const std::string &filename, const std::string &contents); +static gdb::optional gdbpy_colorize_disasm + (const std::string &content, gdbarch *gdbarch); /* The interface between gdb proper and loading of python scripts. */ @@ -162,6 +164,8 @@ static const struct extension_language_ops python_extension_ops = gdbpy_get_matching_xmethod_workers, gdbpy_colorize, + + gdbpy_colorize_disasm, }; /* The main struct describing GDB's interface to the Python @@ -1181,6 +1185,69 @@ gdbpy_colorize (const std::string &filename, const std::string &contents) return std::string (PyBytes_AsString (host_str.get ())); } +/* This is the extension_language_ops.colorize_disasm "method". */ + +static gdb::optional +gdbpy_colorize_disasm (const std::string &content, gdbarch *gdbarch) +{ + if (!gdb_python_initialized) + return {}; + + gdbpy_enter enter_py (get_current_arch (), current_language); + + if (gdb_python_module == nullptr + || !PyObject_HasAttrString (gdb_python_module, "colorize_disasm")) + return {}; + + gdbpy_ref<> hook (PyObject_GetAttrString (gdb_python_module, + "colorize_disasm")); + if (hook == nullptr) + { + gdbpy_print_stack (); + return {}; + } + + if (!PyCallable_Check (hook.get ())) + return {}; + + gdbpy_ref<> content_arg (PyString_FromString (content.c_str ())); + if (content_arg == nullptr) + { + gdbpy_print_stack (); + return {}; + } + + gdbpy_ref<> gdbarch_arg (gdbarch_to_arch_object (gdbarch)); + if (gdbarch_arg == nullptr) + { + gdbpy_print_stack (); + return {}; + } + + gdbpy_ref<> result (PyObject_CallFunctionObjArgs (hook.get (), + content_arg.get (), + gdbarch_arg.get (), + nullptr)); + if (result == nullptr) + { + gdbpy_print_stack (); + return {}; + } + + if (result == Py_None) + return {}; + + gdb::unique_xmalloc_ptr str + = python_string_to_host_string (result.get ()); + if (str == nullptr) + { + gdbpy_print_stack (); + return {}; + } + + return std::string (str.get ()); +} + /* Printing. */ diff --git a/gdb/testsuite/gdb.base/style.exp b/gdb/testsuite/gdb.base/style.exp index 91d3059612d..4cf367ec82d 100644 --- a/gdb/testsuite/gdb.base/style.exp +++ b/gdb/testsuite/gdb.base/style.exp @@ -182,12 +182,25 @@ proc run_style_tests { } { gdb_test_no_output "set width 0" - set main [limited_style main function] - set func [limited_style some_called_function function] - # Somewhere should see the call to the function. - gdb_test "disassemble main" \ - [concat "Dump of assembler code for function $main:.*" \ - "[limited_style $hex address].*$func.*"] + # Disassembly highlighting is done by Python, so, if the + # required modules are not available we'll not get the full + # highlighting. + if { $::python_disassembly_highlighting } { + # Check that the header line of the disassembly output is + # styled correctly, the address at the start of the first + # disassembly line is styled correctly, and that there is at + # least one escape sequence in the disassembly output. + set main [limited_style main function] + gdb_test "disassemble main" \ + [concat "Dump of assembler code for function $main:\\r\\n" \ + "\\s+[limited_style $hex address]\\s+<\\+$decimal>:\[^\\r\\n\]+\033\\\[${decimal}\[^\\r\\n\]+.*" ""] + } else { + set main [limited_style main function] + # Somewhere should see the call to the function. + gdb_test "disassemble main" \ + [concat "Dump of assembler code for function $main:.*" \ + "[limited_style $hex address].*.*"] + } set ifield [limited_style int_field variable] set sfield [limited_style string_field variable] @@ -312,6 +325,25 @@ proc test_startup_version_string { } { gdb_test "" "${vers}.*" "version is styled at startup" } +# Check to see if the Python highlighting of disassembler output is +# expected or not, this highlighting requires Python support in GDB, +# and the Python pygments module to be available. +clean_restart ${binfile} +if {![skip_python_tests]} { + gdb_test_multiple "python import pygments" "" { + -re "ModuleNotFoundError: No module named 'pygments'.*$gdb_prompt $" { + set python_disassembly_highlighting false + } + -re "ImportError: No module named pygments.*$gdb_prompt $" { + set python_disassembly_highlighting false + } + -re "^python import pygments\r\n$gdb_prompt $" { + set python_disassembly_highlighting true + } + } +} else { + set python_disassembly_highlighting false +} # Run tests with all styles in their default state. with_test_prefix "all styles enabled" { -- 2.25.4