From: Andrew Burgess <andrew.burgess@embecosm.com>
To: Tom Tromey <tromey@adacore.com>
Cc: gdb-patches@sourceware.org
Subject: Re: [PATCH] Add some error checking to DWARF assembler
Date: Thu, 18 Mar 2021 17:15:55 +0000 [thread overview]
Message-ID: <20210318171555.GF5520@embecosm.com> (raw)
In-Reply-To: <20210318155826.3703918-1-tromey@adacore.com>
* Tom Tromey <tromey@adacore.com> [2021-03-18 09:58:26 -0600]:
> I had written a DWARF location expression like
>
> DW_OP_const1u
> DW_OP_stack_value
>
> ... and was surprised to see that the DW_OP_stack_value didn't appear
> in the "readelf" output.
>
> The problem here is that DW_OP_const1u requires an operand, but
> neither the DWARF assembler nor gas diagnosed this problem.
>
> This patch adds some checking to Dwarf::_location to try to avoid this
> in the future. I removed some existing checks in the name of
> normalization; while the new messages are a bit worse (being
> one-size-fits-all), I think this doesn't matter hugely, since this
> should only be seen for test suite bugs.
>
> gdb/testsuite/ChangeLog
> 2021-03-18 Tom Tromey <tromey@adacore.com>
>
> * lib/dwarf.exp (Dwarf::_check_n): New proc.
> (Dwarf::_location): Use it.
> ---
> gdb/testsuite/ChangeLog | 5 +++++
> gdb/testsuite/lib/dwarf.exp | 33 +++++++++++++++++++++++----------
> 2 files changed, 28 insertions(+), 10 deletions(-)
>
> diff --git a/gdb/testsuite/lib/dwarf.exp b/gdb/testsuite/lib/dwarf.exp
> index f8fbd381810..64b9e263f10 100644
> --- a/gdb/testsuite/lib/dwarf.exp
> +++ b/gdb/testsuite/lib/dwarf.exp
> @@ -890,6 +890,13 @@ namespace eval Dwarf {
> }
> }
>
> + # Require N arguments for the opcode.
> + proc _check_n {line op n} {
> + if {[llength $line] != $n + 1} {
> + error "$op requires $n arguments"
> + }
> + }
> +
> # This is a miniature assembler for location expressions. It is
> # suitable for use in the attributes to a DIE. Its output is
> # prefixed with "=" to make it automatically use DW_FORM_block.
> @@ -926,56 +933,68 @@ namespace eval Dwarf {
>
> switch -exact -- $opcode {
> DW_OP_addr {
> + _check_n $line $opcode 1
> _op .${addr_size}byte [lindex $line 1]
> }
>
> DW_OP_regx {
> + _check_n $line $opcode 1
> _op .uleb128 [lindex $line 1]
> }
>
> DW_OP_pick -
> DW_OP_const1u -
> DW_OP_const1s {
> + _check_n $line $opcode 1
> _op .byte [lindex $line 1]
> }
>
> DW_OP_const2u -
> DW_OP_const2s {
> + _check_n $line $opcode 1
> _op .2byte [lindex $line 1]
> }
>
> DW_OP_const4u -
> DW_OP_const4s {
> + _check_n $line $opcode 1
> _op .4byte [lindex $line 1]
> }
>
> DW_OP_const8u -
> DW_OP_const8s {
> + _check_n $line $opcode 1
> _op .8byte [lindex $line 1]
> }
>
> DW_OP_constu {
> + _check_n $line $opcode 1
> _op .uleb128 [lindex $line 1]
> }
> DW_OP_consts {
> + _check_n $line $opcode 1
> _op .sleb128 [lindex $line 1]
> }
>
> DW_OP_plus_uconst {
> + _check_n $line $opcode 1
> _op .uleb128 [lindex $line 1]
> }
>
> DW_OP_piece {
> + _check_n $line $opcode 1
> _op .uleb128 [lindex $line 1]
> }
>
> DW_OP_bit_piece {
> + _check_n $line $opcode 2
> _op .uleb128 [lindex $line 1]
> _op .uleb128 [lindex $line 2]
> }
>
> DW_OP_skip -
> DW_OP_bra {
> + _check_n $line $opcode 1
> _op .2byte [lindex $line 1]
> }
>
> @@ -1000,9 +1019,7 @@ namespace eval Dwarf {
>
> DW_OP_implicit_pointer -
> DW_OP_GNU_implicit_pointer {
> - if {[llength $line] != 3} {
> - error "usage: $opcode LABEL OFFSET"
> - }
> + _check_n $line $opcode 2
>
> # Here label is a section offset.
> set label [lindex $line 1]
> @@ -1015,9 +1032,7 @@ namespace eval Dwarf {
> }
>
> DW_OP_GNU_variable_value {
> - if {[llength $line] != 2} {
> - error "usage: $opcode LABEL"
> - }
> + _check_n $line $opcode 1
>
> # Here label is a section offset.
> set label [lindex $line 1]
> @@ -1029,14 +1044,12 @@ namespace eval Dwarf {
> }
>
> DW_OP_deref_size {
> - if {[llength $line] != 2} {
> - error "usage: DW_OP_deref_size SIZE"
> - }
> -
> + _check_n $line $opcode 1
Maybe I'm over thinking this, but you could change this to:
_check_n $line $opcode { SIZE }
And for DW_OP_implicit_pointer above you would do:
_check_n $line $opcode { LABEL OFFSET }
Then in _check_n you can use the length of the final list to validate
the argument count, and if its wrong you can still print the helpful
text....
Just a thought.
Either way, more error checking is good.
Thanks,
Andrew
next prev parent reply other threads:[~2021-03-18 17:15 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-03-18 15:58 Tom Tromey
2021-03-18 17:15 ` Andrew Burgess [this message]
2021-03-18 18:22 ` Tom Tromey
2021-03-18 21:15 ` Tom Tromey
2021-03-31 15:24 ` Tom Tromey
2021-03-31 20:07 ` Andrew Burgess
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=20210318171555.GF5520@embecosm.com \
--to=andrew.burgess@embecosm.com \
--cc=gdb-patches@sourceware.org \
--cc=tromey@adacore.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).