From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 11894 invoked by alias); 14 May 2015 12:06:20 -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 11876 invoked by uid 89); 14 May 2015 12:06:18 -0000 Authentication-Results: sourceware.org; auth=none X-Virus-Found: No X-Spam-SWARE-Status: No, score=-1.9 required=5.0 tests=AWL,BAYES_00,KAM_LAZY_DOMAIN_SECURITY,RCVD_IN_DNSWL_LOW autolearn=no version=3.3.2 X-HELO: mail-oi0-f52.google.com Received: from mail-oi0-f52.google.com (HELO mail-oi0-f52.google.com) (209.85.218.52) by sourceware.org (qpsmtpd/0.93/v0.84-503-g423c35a) with (AES128-GCM-SHA256 encrypted) ESMTPS; Thu, 14 May 2015 12:06:17 +0000 Received: by oiko83 with SMTP id o83so53711607oik.1 for ; Thu, 14 May 2015 05:06:15 -0700 (PDT) X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20130820; h=x-gm-message-state:mime-version:in-reply-to:references:from:date :message-id:subject:to:cc:content-type; bh=w1+3GJMr5Y3+Lau4GpMZ+DGxEAbX1ruQZDuiPhSFmLo=; b=QA/CHMhenIdyqGCIq5EOu71sVf8R+lCMaqlZkinbMxUT2Hpdck4vQht726TNXtP3KZ Mcuh6qCV6qhNImLis3IO6fT5hBWKhQh6xRcUMrJ7VrDOqoM4XDNud3cPrPzgjiB4alxX wanDFVAbKSdGiYQg1r7Uxf8Qg0wwsGPPLdIXDRrqwxN4B4qLe6F8X4XCqpmizVUDHgWP WEwR5X45r0CbrE7pJ/Kn4WMwuGtl2+NpDcENadyw5agUXupfxKe29TwJrUldGXvdYaxZ KnxEMFt7aayFTuD1UEQzQ5lLk4C7rYd5iybNAH30pGltSvMqWM/vFfsYQU+W7WPCu6/U zMcQ== X-Gm-Message-State: ALoCoQmFsq6PIwmXPvt73Zlr/Rv/2msaaUyYkgtaHx+r60HTjSxwOrR77QKPFVs6IQ+0jwpqjgO4 X-Received: by 10.202.54.3 with SMTP id d3mr2969520oia.103.1431605175126; Thu, 14 May 2015 05:06:15 -0700 (PDT) MIME-Version: 1.0 Received: by 10.182.243.232 with HTTP; Thu, 14 May 2015 05:05:54 -0700 (PDT) In-Reply-To: <555475F1.2030504@redhat.com> References: <1431555450-15493-1-git-send-email-patrick@parcs.ath.cx> <555475F1.2030504@redhat.com> From: Patrick Palka Date: Thu, 14 May 2015 12:06:00 -0000 Message-ID: Subject: Re: [PATCH] Fix PR gdb/16999 To: Pedro Alves Cc: "gdb-patches@sourceware.org" Content-Type: text/plain; charset=UTF-8 X-SW-Source: 2015-05/txt/msg00364.txt.bz2 On Thu, May 14, 2015 at 6:16 AM, Pedro Alves wrote: > On 05/13/2015 11:17 PM, Patrick Palka wrote: >> When GDB reads a nonsensical value for the HISTSIZE environment variable >> variable, i.e. one that is non-numeric or negative, GDB then sets its >> history size to 0. This behavior is contrary to that of bash, which >> defaults the history size to unlimited in such cases. >> >> This patch makes the behavior of invalid HISTSIZE match that of bash. >> When we encounter an invalid HISTSIZE we now set the history size to >> unlimited instead of 0. >> >> gdb/ChangeLog: >> >> PR gdb/16999 >> * top.c (init_history): For invalid HISTSIZE, set history size >> to unlimited. >> >> gdb/testsuite/ChangeLog: >> >> PR gdb/16999 >> * gdb.base/histsize-history.exp: New test. >> --- >> gdb/testsuite/gdb.base/histsize-history.exp | 54 +++++++++++++++++++++++++++++ >> gdb/top.c | 16 ++++----- >> 2 files changed, 62 insertions(+), 8 deletions(-) >> create mode 100644 gdb/testsuite/gdb.base/histsize-history.exp >> >> diff --git a/gdb/testsuite/gdb.base/histsize-history.exp b/gdb/testsuite/gdb.base/histsize-history.exp >> new file mode 100644 >> index 0000000..b7b13cf >> --- /dev/null >> +++ b/gdb/testsuite/gdb.base/histsize-history.exp >> @@ -0,0 +1,54 @@ >> +# Copyright 2015 Free Software Foundation, Inc. >> + >> +# This program is free software; you can redistribute it and/or modify >> +# it under the terms of the GNU General Public License as published by >> +# the Free Software Foundation; either version 3 of the License, or >> +# (at your option) any later version. >> +# >> +# This program is distributed in the hope that it will be useful, >> +# but WITHOUT ANY WARRANTY; without even the implied warranty of >> +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the >> +# GNU General Public License for more details. >> +# >> +# You should have received a copy of the GNU General Public License >> +# along with this program. If not, see . >> + >> +# This file is part of the gdb testsuite. >> + >> +# Test the setting of "history size" via the HISTSIZE environment variable >> + >> + >> +# Check that the history size is properly set to SIZE when env(HISTSIZE) is set >> +# to HISTSIZE. >> + >> +proc test_histsize_history_setting { histsize size } { >> + global env >> + >> + if [info exists env(HISTSIZE)] { >> + set old_histsize $env(HISTSIZE) >> + } >> + set env(HISTSIZE) $histsize >> + >> + gdb_exit >> + gdb_start >> + >> + gdb_test "show history size" "The size of the command history is $size." >> + >> + if { $size == "0" } { >> + gdb_test_no_output "show commands" >> + } elseif { $size != "1" } { >> + gdb_test "show commands" " . show history size\r\n . show commands" >> + } >> + >> + if [info exists old_histsize] { >> + set env(HISTSIZE) $old_histsize >> + } else { >> + unset env(HISTSIZE) >> + } >> +} >> + >> +test_histsize_history_setting "0" "0" >> +test_histsize_history_setting "20" "20" >> +test_histsize_history_setting "-5" "unlimited" >> +test_histsize_history_setting "not_an_integer" "unlimited" >> +test_histsize_history_setting "10zab" "unlimited" >> diff --git a/gdb/top.c b/gdb/top.c >> index 74e1e07..00fee8d 100644 >> --- a/gdb/top.c >> +++ b/gdb/top.c >> @@ -1684,15 +1684,15 @@ init_history (void) >> if (tmpenv) >> { >> int var; >> + char *endptr; >> >> - var = atoi (tmpenv); >> - if (var < 0) >> - { >> - /* Prefer ending up with no history rather than overflowing >> - readline's history interface, which uses signed 'int' >> - everywhere. */ >> - var = 0; >> - } >> + var = strtol (tmpenv, &endptr, 10); >> + >> + /* For the sake of consistency with bash, if HISTSIZE is >> + non-numeric or if HISTSIZE is negative then set our history >> + size to unlimited. */ >> + if (*endptr != '\0' || var < 0) >> + var = -1; > > Hmm, looking at the master branch (Bash-4.3 patch 33) in: > http://git.savannah.gnu.org/cgit/bash.git > > This seems to be where bash implements this: > > #if defined (HISTORY) > /* What to do after the HISTSIZE or HISTFILESIZE variables change. > If there is a value for this HISTSIZE (and it is numeric), then stifle > the history. Otherwise, if there is NO value for this variable, > ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ > unstifle the history. If name is HISTFILESIZE, and its value is > ^^^^^^^^^^^^^^^^^^^^ > numeric, truncate the history file to hold no more than that many > lines. */ > void > sv_histsize (name) > char *name; > { > char *temp; > intmax_t num; > int hmax; > > temp = get_string_value (name); > > if (temp && *temp) > { > if (legal_number (temp, &num)) > { > hmax = num; > if (hmax < 0 && name[4] == 'S') > unstifle_history (); /* unstifle history if HISTSIZE < 0 */ > else if (name[4] == 'S') > { > stifle_history (hmax); > hmax = where_history (); > if (history_lines_this_session > hmax) > history_lines_this_session = hmax; > } > else if (hmax >= 0) /* truncate HISTFILE if HISTFILESIZE >= 0 */ > { > history_truncate_file (get_string_value ("HISTFILE"), hmax); > if (hmax <= history_lines_in_file) > history_lines_in_file = hmax; > } > } > } > else if (name[4] == 'S') > unstifle_history (); > } > > > Note the comment I underlined above. > > > If I'm reading right, stripping out all the HISTFILESIZE handling > and then removing dead code, we get: > > void > sv_histsize (char * name) > { > char *temp; > intmax_t num; > int hmax; > > temp = get_string_value (name); > > if (temp && *temp) > { > if (legal_number (temp, &num)) > { > hmax = num; > if (hmax < 0) > unstifle_history (); /* unstifle history if HISTSIZE < 0 */ > else > stifle_history (hmax); > } > } > else > unstifle_history (); > } > > Note: > > #1 - if HISTSIZE is non-numeric nothing happens. I think that means > bash ends up with the default history size. > > #2 - if HISTSIZE is set to the empty string, bash ends up > with unlimited history size. > > It seems to me that your patch handles both of these differently. > > #2 appears consistent with what is suggested here: > > http://stackoverflow.com/questions/9457233/unlimited-bash-history > > "Set HISTSIZE and HISTFILESIZE to an empty string: > > HISTSIZE= HISTFILESIZE= > > In bash 4.3 and later you can also use HISTSIZE=-1 HISTFILESIZE=-1: > " When HISTSIZE is the empty string, I think we invoke UB. I'm not sure atoi() I was basing this patch off of the behavior of HISTFILESIZE, not HISTSIZE. This is because the behavior of HISTFILESIZE is more specified than that of HISTSIZE. According to the documentation for HISTFILESIZE: Non-numeric values and numeric values less than zero inhibit truncation. The shell sets the default value to the value of HISTSIZE after reading any startup files. ... If HISTFILESIZE is unset, or set to null, a non-numeric value, or a numeric value less than zero, the history file is not truncated. Whereas for HISTSIZE, the documentation just says: If the value is 0, commands are not saved in the history list. Numeric values less than zero result in every command being saved on the history list (there is no limit). bash does not explicitly document what happens if HISTSIZE is null or non-numeric. And because HISTSIZE determines both GDB's in-memory and in-file history size (we don't respond to HISTFILESIZE) I think it's not so bad to base our behavior of HISTSIZE handling off of bash's HISTFILESIZE handling. (Though I should've mentioned this in the patch.) But unless GDB responds to both HISTSIZE and HISTFILESIZE I suppose we will never be truly consistent with bash since apparently both of these variables have different behaviors under edge cases. So should we anyway match the behavior of bash's HISTSIZE?