Message ID | 5a785bba-7432-f6e0-1089-5d2bdd3450a3@gmail.com |
---|---|
State | New, archived |
Headers |
Received: (qmail 27092 invoked by alias); 28 Feb 2019 12:52:30 -0000 Mailing-List: contact gdb-patches-help@sourceware.org; run by ezmlm Precedence: bulk List-Id: <gdb-patches.sourceware.org> List-Unsubscribe: <mailto:gdb-patches-unsubscribe-##L=##H@sourceware.org> List-Subscribe: <mailto:gdb-patches-subscribe@sourceware.org> List-Archive: <http://sourceware.org/ml/gdb-patches/> List-Post: <mailto:gdb-patches@sourceware.org> List-Help: <mailto:gdb-patches-help@sourceware.org>, <http://sourceware.org/ml/#faqs> Sender: gdb-patches-owner@sourceware.org Delivered-To: mailing list gdb-patches@sourceware.org Received: (qmail 26692 invoked by uid 89); 28 Feb 2019 12:52:30 -0000 Authentication-Results: sourceware.org; auth=none X-Spam-SWARE-Status: No, score=-22.6 required=5.0 tests=BAYES_00, BODY_8BITS, FREEMAIL_ENVFROM_END_DIGIT, FREEMAIL_FROM, GARBLED_BODY, GIT_PATCH_0, GIT_PATCH_1, GIT_PATCH_2, GIT_PATCH_3, RCVD_IN_DNSWL_NONE, SPF_PASS autolearn=ham version=3.3.2 spammy=2017-02, 201702, H*RU:sk:broadba, Hx-spam-relays-external:sk:broadba X-HELO: mail-lf1-f68.google.com Received: from mail-lf1-f68.google.com (HELO mail-lf1-f68.google.com) (209.85.167.68) by sourceware.org (qpsmtpd/0.93/v0.84-503-g423c35a) with ESMTP; Thu, 28 Feb 2019 12:52:28 +0000 Received: by mail-lf1-f68.google.com with SMTP id a8so9357105lfi.7 for <gdb-patches@sourceware.org>; Thu, 28 Feb 2019 04:52:27 -0800 (PST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20161025; h=to:from:autocrypt:subject:message-id:date:user-agent:mime-version; bh=8kIypNxvgHa8nmc38IMgEpfuSDmWKne8ppT6XrWM1H0=; b=UH1ezxWEkklpfOGwVq/WfpF64oLcj2rQupiAMfNq2xNsHa2h8Bc30jki+CjCWkieDz Y5wP5wP93JC4Vd7b2PDOFkp2xhSKVN3ashIWDIbueoYa/zlGwXbQduUSBMXblnV6yxDs g53txKLmmjrT6nXygLd9bpkBERDC2TveGajnPxdvPAzBgtCsunmS2nXd7l5LxD6q1FtT 6VH/ozpOvAAtgpsHapIakJQhAcdD7yynmTRkK9YSNPQ93xHbFgIE/st7SR8wJLdFZc8S DqO31z7uFM2pK2Pf7jt4bp3+cIWAxU3ymDL/FiV6/YL8T+Te0Ps4CoNhBPPSYsZS/TDu GA/w== Return-Path: <lrn1986@gmail.com> Received: from [192.168.4.39] (broadband-95-84-200-6.ip.moscow.rt.ru. [95.84.200.6]) by smtp.gmail.com with ESMTPSA id e23sm4102582ljj.3.2019.02.28.04.52.23 for <gdb-patches@sourceware.org> (version=TLS1_2 cipher=ECDHE-RSA-AES128-GCM-SHA256 bits=128/128); Thu, 28 Feb 2019 04:52:23 -0800 (PST) To: gdb-patches@sourceware.org From: LRN <lrn1986@gmail.com> Subject: [PATCH] Apply substitute-path to relative filenames as well Message-ID: <5a785bba-7432-f6e0-1089-5d2bdd3450a3@gmail.com> Date: Thu, 28 Feb 2019 15:52:15 +0300 User-Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:60.0) Gecko/20100101 Thunderbird/60.5.1 MIME-Version: 1.0 Content-Type: multipart/signed; micalg=pgp-sha256; protocol="application/pgp-signature"; boundary="vn67TU5XTP09kobpkjLx4RhQNT3N2oDhL" X-IsSubscribed: yes |
Commit Message
LRN
Feb. 28, 2019, 12:52 p.m. UTC
The patch is attached.
The change itself is similar to
https://www.sourceware.org/ml/gdb-patches/2017-02/msg00693.html , but the code
is slightly cleaner (at the cost of making the patch larger).
Also, as the patch is trivial, so i would expect it to not to require copyright
assignment (i do have one, but David Grayson might not have).
Here's the copy of the commit message:
When source file path is relative to the build directory (which
is considered a good practice and is enforced in certain buildsystems,
such as meson), gdb only applies substitute-path to the build directory
path. Then gdb appends the source file path to the rewritten build
directory path, and tries to access that.
This fails if either two of the following conditions are true:
a) The user didn't specify substitute-path for the build directory.
This is highly likely, since path substitution for build directories
is not documented anywhere, and since gdb does not tell[0] the user
the path to the build directory, just the source file path.
b) The source file path changed.
This can also easily happen, since a source path that is relative
to the build directory can include any number of directory names
that are not part of the program source tree (starting with the
name of the root directory of the source tree). Gdb will not apply
substitute-path to that relative path, thus there is no way for
the user to tell gdb about these changes.
This commit changes the code to apply substitute-path to all filenames,
both relative and absolute. This way it is possible to do things like:
set substitute-path ../foobar-1.0 /src/my/foobar-1.0
which is completely in line with the user expectations.
This might break unusual cases where build directory path is also
relative (is that even possible?) and happens to match the path
to the source directory (i.e. happens to match a substitution rule).
[0]: There's a "maintenance info symtabs" command that does show the names
of the build directories, but normal users are not required to
know or use that.
---
gdb/source.c | 12 ++++--------
1 file changed, 4 insertions(+), 8 deletions(-)
gdb/ChangeLog:
2019-02-28 Руслан Ижбулатов <lrn1986@gmail.com>
* source.c (find_and_open_source): Apply substitute-path to all
filenames, both absolute and relative
From 28431cdf98ec6c606373fd02c36c5642c4f196df Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=D0=A0=D1=83=D1=81=D0=BB=D0=B0=D0=BD=20=D0=98=D0=B6=D0=B1?=
=?UTF-8?q?=D1=83=D0=BB=D0=B0=D1=82=D0=BE=D0=B2?= <lrn1986@gmail.com>
Date: Thu, 28 Feb 2019 10:25:41 +0000
Subject: [PATCH] Apply substitute-path to relative filenames as well
When source file path is relative to the build directory (which
is considered a good practice and is enforced in certain buildsystems,
such as meson), gdb only applies substitute-path to the build directory
path. Then gdb appends the source file path to the rewritten build
directory path, and tries to access that.
This fails if either two of the following conditions are true:
a) The user didn't specify substitute-path for the build directory.
This is highly likely, since path substitution for build directories
is not documented anywhere, and since gdb does not tell[0] the user
the path to the build directory, just the source file path.
b) The source file path changed.
This can also easily happen, since a source path that is relative
to the build directory can include any number of directory names
that are not part of the program source tree (starting with the
name of the root directory of the source tree). Gdb will not apply
substitute-path to that relative path, thus there is no way for
the user to tell gdb about these changes.
This commit changes the code to apply substitute-path to all filenames,
both relative and absolute. This way it is possible to do things like:
set substitute-path ../foobar-1.0 /src/my/foobar-1.0
which is completely in line with the user expectations.
This might break unusual cases where build directory path is also
relative (is that even possible?) and happens to match the path
to the source directory (i.e. happens to match a substitution rule).
[0]: There's a "maintenance info symtabs" command that does show the names
of the build directories, but normal users are not required to
know or use that.
---
gdb/source.c | 12 ++++--------
1 file changed, 4 insertions(+), 8 deletions(-)
Comments
>>>>> "LRN" == LRN <lrn1986@gmail.com> writes:
LRN> The patch is attached.
LRN> The change itself is similar to
LRN> https://www.sourceware.org/ml/gdb-patches/2017-02/msg00693.html , but the code
LRN> is slightly cleaner (at the cost of making the patch larger).
Thank you for the patch.
LRN> Also, as the patch is trivial, so i would expect it to not to require copyright
LRN> assignment (i do have one, but David Grayson might not have).
I agree.
LRN> 2019-02-28 Руслан Ижбулатов <lrn1986@gmail.com>
LRN> * source.c (find_and_open_source): Apply substitute-path to all
LRN> filenames, both absolute and relative
Is it possible to get a test case for the patch?
LRN> gdb::unique_xmalloc_ptr<char> rewritten_filename;
[...]
LRN> + rewritten_filename = rewrite_source_path (filename);
I think these two lines can now be joined; no need to have a separate
assignment.
thanks,
Tom
On 06.03.2019 1:10, Tom Tromey wrote: > Thank you for the patch. > Hey, it's been a week. Do i need to do something to move this along? (Please, don't say "test case" :( )
On 13.03.2019 15:47, LRN wrote: > On 06.03.2019 1:10, Tom Tromey wrote: >> Thank you for the patch. >> > > Hey, it's been a week. Do i need to do something to move this along? (Please, > don't say "test case" :( ) > Hey, it's been two weeks. Do i need to do something to move this along? (Please, don't say "test case" :( )
>>>>> "LRN" == LRN <lrn1986@gmail.com> writes: >> Thank you for the patch. LRN> Hey, it's been a week. Do i need to do something to move this along? (Please, LRN> don't say "test case" :( ) Is it possible to write one? I didn't think about it in detail. The norm is that a patch should come with a test case. Exceptions are for refactoring patches, where the intent is to preserve semantics; and for the relatively rare cases where a test is very difficult or impossible to write. I see there are already tests in gdb.base/subst.exp. Perhaps it could be extended? Also the review had a little nit that could be addressed. thanks, Tom
On 28.03.2019 0:36, Tom Tromey wrote: > LRN writes: > >> Hey, it's been a week. Do i need to do something to move this along? (Please, >> don't say "test case" :( ) > > Is it possible to write one? I didn't think about it in detail. > > I see there are already tests in gdb.base/subst.exp. Perhaps it could > be extended? These tests only test the interactive output of gdb with regards to subst commands. I.e. it issues a command, such as "show substitute-path", and then matches the output to the one it expects. My patch influences the output only when gdb is actually debugging something - which means that the test must compile some program (using relative path to the source file!), run it under gdb, set up a breakpoint, verify that the source path printed at the breakpoint hit is relative (and unresolved - the test would probably need to move the source file before running?), then use subst, then probably restart and verify that the path is non-relative and resolved when the same breakpoint hits. Okay, maybe it can be done without breakpoints - just run and list ../../path/to/sourcefile.c:0 or something. Anyway, that still sounds an order of magnitude more complicated that subst.exp, so i would rather not try to implement all that using completely unfamiliar testing framework that i do not have. > > Also the review had a little nit that could be addressed. > I did address it. The first follow-up[0] message had the new version of the patch attached, with the nit fixed. [0]: https://www.sourceware.org/ml/gdb-patches/2019-03/msg00082.html
>>>>> "LRN" == LRN <lrn1986@gmail.com> writes:
LRN> Anyway, that still sounds an order of magnitude more complicated that
LRN> subst.exp, so i would rather not try to implement all that using completely
LRN> unfamiliar testing framework that i do not have.
Alright, I will see if I can write one.
LRN> I did address it. The first follow-up[0] message had the new version of the
LRN> patch attached, with the nit fixed.
Thanks, I must have missed that.
Tom
>>>>> "Tom" == Tom Tromey <tom@tromey.com> writes: >>>>> "LRN" == LRN <lrn1986@gmail.com> writes: LRN> Anyway, that still sounds an order of magnitude more complicated that LRN> subst.exp, so i would rather not try to implement all that using completely LRN> unfamiliar testing framework that i do not have. Tom> Alright, I will see if I can write one. LRN> I did address it. The first follow-up[0] message had the new version of the LRN> patch attached, with the nit fixed. Tom> Thanks, I must have missed that. I'm checking this in. I finally looked, and writing a test did seem like a pain. Tom
diff --git a/gdb/source.c b/gdb/source.c index f99215f..673ebc3 100644 --- a/gdb/source.c +++ b/gdb/source.c @@ -1028,15 +1028,11 @@ find_and_open_source (const char *filename, } gdb::unique_xmalloc_ptr<char> rewritten_filename; - if (IS_ABSOLUTE_PATH (filename)) - { - /* If filename is absolute path, try the source path - substitution on it. */ - rewritten_filename = rewrite_source_path (filename); - if (rewritten_filename != NULL) - filename = rewritten_filename.get (); - } + rewritten_filename = rewrite_source_path (filename); + + if (rewritten_filename != NULL) + filename = rewritten_filename.get (); result = openp (path, OPF_SEARCH_IN_PATH | OPF_RETURN_REALPATH, filename, OPEN_MODE, fullname);