This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [RFA] linux-nat.c minor cleanup
- From: Doug Evans <dje at google dot com>
- To: gdb-patches at sourceware dot org
- Date: Fri, 13 Mar 2009 10:10:06 -0700
- Subject: Re: [RFA] linux-nat.c minor cleanup
- References: <20081219053421.2D802412300@localhost> <e394668d0901051504g3105e2b0p533d9f983b05af83@mail.gmail.com> <e394668d0901051600w109d4cc9l800c68c56e5d52af@mail.gmail.com>
Ping.
On Mon, Jan 5, 2009 at 5:00 PM, Doug Evans <dje@google.com> wrote:
> On Mon, Jan 5, 2009 at 3:04 PM, Doug Evans <dje@google.com> wrote:
>> Ping.
>>
>> On Thu, Dec 18, 2008 at 9:34 PM, Doug Evans <dje@google.com> wrote:
>>> Hi.
>>>
>>> linux-nat.c:linux_nat_info_proc_cmd uses a long long to record a pid.
>>> There's not much point in that, so this patch changes it to a long.
>>> [either that or it shouldn't use strtoul to parse it :-)]
>>>
>>> The handling of info proc cmd,cwd,exe is odd too.
>>> They're always printed and yet argument parsing looks for them anyway,
>>> and ignores invalid commands instead of flagging an error.
>>> This patch adds a blurb about printing cmd,cwd,exe in the help text.
>>> The manual already touches on this, though it doesn't precisely
>>> say that cmd,cwd,exe are always printed.
>>> I wouldn't mind also submitting a patch to flag unknown commands
>>> as errors, but I'm not sure what's intended as far as the parsing
>>> of cmd,cwd,exe.
>
> Going over the file I found a case I missed.
> atoi has undefined behaviour for invalid input so I stuck with strtoul
> in linux_nat_info_proc_cmd. Granted, it is awkward that some fns use
> long and some fns use int, but ptid_get_pid returns an int and strotul
> returns a long. strtopid anyone?
>
> [I realize this is just minor cleanup, but for completeness' sake,
> there are several uses of pid_t in linux-nat.c - maybe pid_t should be
> used everywhere instead of int - if one wants to be pedantic. But
> then one is left with either assuming it's an int or adding int casts
> to sprintf arguments (which the code already does). I don't have a
> preference, I'll make the patch look like whatever y'all want. It's
> just annoying to see long long used here.]
>
> This version removes the change to the doc string for "info proc". I
> could still use some input on what preferred: remove the argument
> processing to watch for cmd, cwd, exe?
>
> 2008-12-18 Doug Evans <dje@google.com>
>
> * linux-nat.c (linux_nat_find_memory_regions): Result of PIDGET is an
> int, not a long long.
> (linux_nat_info_proc_cmd): Store pid in long instead of long long.
>