From: Simon Marchi <simon.marchi@polymtl.ca>
To: Pedro Alves <palves@redhat.com>
Cc: gdb-patches@sourceware.org
Subject: Re: [PATCH v4 1/2] Add back gdb_pretty_print_insn
Date: Wed, 01 Feb 2017 18:10:00 -0000 [thread overview]
Message-ID: <df4b6203d89992b02f45dae5c45a2fa4@polymtl.ca> (raw)
In-Reply-To: <1485909045-30285-2-git-send-email-palves@redhat.com>
On 2017-01-31 19:30, Pedro Alves wrote:
> ui_file_rewind is a ui_file method that only really works with mem
> buffer files, and is a nop on other ui_file types. It'd be desirable
> to eliminate it from the base ui_file interface, and move it to the
> "mem_fileopen" subclass of ui_file instead. A following patch does
> just that.
>
> Unfortunately, there are a couple references to ui_file_rewind inside
> gdb_disassembler::pretty_print_insn that were made harder to eliminate
> with the recent addition of the gdb_disassembler wrapper.
>
> Before the gdb_disassembler wrapper was added, in commit
> e47ad6c0bd7aa3 ("Refactor disassembly code"), gdb_pretty_print_insn
> used to be passed a ui_file pointer as argument, and it was simple to
> adjust that pointer be a "mem_fileopen" ui_file pointer instead, since
> there's only one gdb_pretty_print_insn caller.
>
> That commit made gdb_pretty_print_insn be a method of
> gdb_disassembler, and removed the method's ui_file parameter at the
> same time, replaced by referencing the gdb_disassembler's stream
> instead. The trouble is that a gdb_disassembler can be instantiated
> with a pointer any kind of ui_file. Casting the gdb_disassembler's
> stream to a mem_fileopen ui_file inside
> gdb_disassembler::pretty_print_insn in order to call the reset method
> would be gross hack.
>
> The fix here is to:
>
> - make gdb_disassembler::pretty_print_insn a be free function again
> instead of a method of gdb_disassembler. I.e., bring back
> gdb_pretty_print_insn.
>
> - but, don't add back the ui_file * parameter. We'd always be
> passing in a fresh mem_fileopen anyway, so move the mem_fileopen
> allocation inside. That is a better interface, given that the
> ui_file is only ever used as temporary scratch buffer as an
> implementation detail of gdb_pretty_print_insn. The function's
> real "where to send output" parameter is the ui_out pointer.
>
> - don't add back a disassemble_info pointer either. That used to be
> necessary for this bit:
>
> err = m_di.read_memory_func (pc, &data, 1, &m_di);
> if (err != 0)
> m_di.memory_error_func (err, pc, &m_di);
>
> ... but AFAIK, it's not really necessary. We can replace those
> three lines with a call to read_code. This seems to fix a
> regression even, because before commit d8b49cf0c891d0 ("Don't throw
> exception in dis_asm_memory_error"), that memory_error_func call
> would throw an error/exception, but now it only records the error
> in the gdb_disassembler's m_err_memaddr field. (read_code throws
> on error.)
>
> With all these, gdb_pretty_print_insn is completely layered on top of
> gdb_disassembler only using the latter's public API.
I don't think I understand the situation fully, but what you suggest
looks good to me. I was confused by the fact that the gdb_disassembler
constructor accepts a stream, but the pretty_print_insn method takes a
uiout. Which one is used for printing then? I think that your patch
clears that up.
The only possible issue I can see is that in your version, one
gdb_disassembler and one string_file object are constructed for each
disassembled instruction, rather than re-using them for as long as we
need to disassemble. I don't know how much impact it has on the
performance (probably negligible), but something to keep in mind.
next prev parent reply other threads:[~2017-02-01 18:10 UTC|newest]
Thread overview: 18+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-02-01 0:30 [PATCH v4 0/2] Eliminate cleanups & make ui_file a C++ class hierarchy Pedro Alves
2017-02-01 0:30 ` [PATCH v4 1/2] Add back gdb_pretty_print_insn Pedro Alves
2017-02-01 17:01 ` Luis Machado
2017-02-01 18:10 ` Simon Marchi [this message]
2017-02-01 18:26 ` Simon Marchi
2017-02-02 0:00 ` Pedro Alves
2017-02-01 20:02 ` Pedro Alves
2017-02-01 20:31 ` Pedro Alves
2017-02-01 23:50 ` Pedro Alves
2017-02-02 1:20 ` Simon Marchi
2017-02-02 11:37 ` [pushed] Reuse buffers across gdb_pretty_print_insn calls (Re: [PATCH v4 1/2] Add back gdb_pretty_print_insn) Pedro Alves
2017-02-01 0:31 ` [PATCH v4 2/2] Eliminate make_cleanup_ui_file_delete / make ui_file a class hierarchy Pedro Alves
2017-02-01 17:37 ` Luis Machado
2017-02-01 22:49 ` Pedro Alves
2017-02-01 23:24 ` Luis Machado
2017-02-02 0:02 ` Pedro Alves
2017-02-27 19:43 ` Edjunior Barbosa Machado
2017-03-07 14:02 ` Pedro Alves
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=df4b6203d89992b02f45dae5c45a2fa4@polymtl.ca \
--to=simon.marchi@polymtl.ca \
--cc=gdb-patches@sourceware.org \
--cc=palves@redhat.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).