From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 6059 invoked by alias); 28 Jan 2019 20:47:17 -0000 Mailing-List: contact gdb-patches-help@sourceware.org; run by ezmlm Precedence: bulk List-Id: List-Subscribe: List-Archive: List-Post: List-Help: , Sender: gdb-patches-owner@sourceware.org Received: (qmail 6044 invoked by uid 89); 28 Jan 2019 20:47:17 -0000 Authentication-Results: sourceware.org; auth=none X-Spam-SWARE-Status: No, score=-11.2 required=5.0 tests=BAYES_00,GIT_PATCH_2,GIT_PATCH_3,SPF_HELO_PASS,SPF_SOFTFAIL autolearn=ham version=3.3.2 spammy=175, widespread, HContent-Transfer-Encoding:8bit X-HELO: mail.baldwin.cx Received: from bigwig.baldwin.cx (HELO mail.baldwin.cx) (96.47.65.170) by sourceware.org (qpsmtpd/0.93/v0.84-503-g423c35a) with ESMTP; Mon, 28 Jan 2019 20:47:14 +0000 Received: from ralph.baldwin.cx (ralph.baldwin.cx [66.234.199.215]) by mail.baldwin.cx (Postfix) with ESMTPSA id B566F10B429 for ; Mon, 28 Jan 2019 15:47:12 -0500 (EST) From: John Baldwin To: gdb-patches@sourceware.org Subject: [PATCH v2 0/4] Some fixes for debug files and sysroots Date: Mon, 28 Jan 2019 20:47:00 -0000 Message-Id: MIME-Version: 1.0 Content-Transfer-Encoding: 8bit X-IsSubscribed: yes X-SW-Source: 2019-01/txt/msg00589.txt.bz2 Relative to the first series: 1) I combined the duplicate checks for "are we in a sysroot" in the first patch as as suggested by Simon. 2) I dropped the second patch (trim trailing '/' from sysroot). 3) Patches 2 and 3 are a different take on solving the issue when the sysroot ends in '/'. Patch 2 adds a 'child_path' function to determine if a child path is a child of a parent (requiring the child to have at least one component "below" the parent). It also returns a pointer to the first component below the parent (but after the directory separator). Patch 3 uses child_path in find_separate_debug_file which fixes it in the case that the sysroot ends in a /. 4) The 4th patch is a new patch for a different issue I ran into while testing this some more today. The paths to object files are always canonical paths with symlinks resolved. If the sysroot entered by the user is a path containing symlinks, the filename_ncmp will never match. To handle sysroot paths that traverse symlinks, use gdb_realpath to generate a canonical sysroot path and use that instead of gdb_sysroot with child_path. As an aside, it's not clear to me when one should use gdb_realpath instead of lrealpath. gdb_realpath seems more widespread and also returns an RAII-friendly type, so I used that. John Baldwin (4): Look for separate debug files in debug directories under a sysroot. Add a new function child_path. Use child_path to determine if an object file is under a sysroot. Try to use the canonical version of a sysroot for debug file links. gdb/ChangeLog | 23 ++++++++++ gdb/Makefile.in | 1 + gdb/common/pathstuff.c | 52 ++++++++++++++++++++++ gdb/common/pathstuff.h | 6 +++ gdb/symfile.c | 36 +++++++++++++--- gdb/unittests/child-path-selftests.c | 64 ++++++++++++++++++++++++++++ 6 files changed, 175 insertions(+), 7 deletions(-) create mode 100644 gdb/unittests/child-path-selftests.c -- 2.19.2