Patchwork cc-with-tweaks: show dwz stderr and check exit code

login
register
mail settings
Submitter Simon Marchi
Date May 8, 2019, 4 p.m.
Message ID <20190508160012.32596-1-simon.marchi@efficios.com>
Download mbox | patch
Permalink /patch/32608/
State New
Headers show

Comments

Simon Marchi - May 8, 2019, 4 p.m.
When running the gdb.base/index-cache.exp test case with the
cc-with-dwz-m board, I noticed that the final executable didn't actually
contain a .gnu_debugaltlink section with the name of the external dwz
file:

    $ readelf --debug-dump=links testsuite/outputs/gdb.base/index-cache/index-cache
    * empty *

Running dwz by hand, I realized it's because dwz complains that the
output .debug_info section is empty and fails:

    $ gcc ~/src/binutils-gdb/gdb/testsuite/gdb.base/index-cache.c -g3 -O0 -o a && cp a b
    $ dwz -m foo a b
    dwz: foo: .debug_info section not present
    $ echo $?
    1

This is because index-cache.c is trivial (just an empty main) and dwz
doesn't find anything to factor out to the dwz file. [1]

I think that cc-with-tweaks should fail in this scenario: if the user
asks for an external dwz file to be generated (the -m flag), then it
should be an error if cc-with-tweaks doesn't manage to produce an
executable with the proper link to this external dwz file.  Otherwise,
the test runs with a regular non-dwzified executable, which gives a
false sense of security about whether the feature under test works with
dwzified executables.

So this patch adds checks for that after invoking dwz.  It also removes
the 2>&1 to allow the error message to be printed like so:

    Running /home/smarchi/src/binutils-gdb/gdb/testsuite/gdb.base/index-cache.exp ...
    gdb compile failed, dwz: /home/smarchi/build/binutils-gdb/gdb/testsuite/outputs/gdb.base/index-cache/index-cache.dwz: .debug_info section not present

With this patch, fewer tests will pass than before with the
cc-with-dwz and cc-with-dwz-m boards, but those were false positives
anyway.

[1] Note that dwz has been patched by Tom de Vries to work correctly in
this case, so we can use dwz master to run the test:

https://sourceware.org/git/?p=dwz.git;a=commit;h=08becc8b33453b6d013a65e7eeae57fc1881e801

gdb/ChangeLog:

	* contrib/cc-with-tweaks.sh: Check for return value of dwz.
---
 gdb/contrib/cc-with-tweaks.sh | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)
Tom de Vries - May 9, 2019, 11:25 a.m.
On 08-05-19 18:00, Simon Marchi wrote:
> When running the gdb.base/index-cache.exp test case with the
> cc-with-dwz-m board, I noticed that the final executable didn't actually
> contain a .gnu_debugaltlink section with the name of the external dwz
> file:
> 
>     $ readelf --debug-dump=links testsuite/outputs/gdb.base/index-cache/index-cache
>     * empty *
> 
> Running dwz by hand, I realized it's because dwz complains that the
> output .debug_info section is empty and fails:
> 
>     $ gcc ~/src/binutils-gdb/gdb/testsuite/gdb.base/index-cache.c -g3 -O0 -o a && cp a b
>     $ dwz -m foo a b
>     dwz: foo: .debug_info section not present
>     $ echo $?
>     1
> 
> This is because index-cache.c is trivial (just an empty main) and dwz
> doesn't find anything to factor out to the dwz file. [1]
> 
> I think that cc-with-tweaks should fail in this scenario:


> if the user
> asks for an external dwz file to be generated (the -m flag),

This dwz semantics is open for debate, I think.

In the case of dwz a, the user asks for a to be compressed, and if that
doesn't happen due to size heuristics, then execution is still
considered a success.

In the case of dwz a -o b, the user asks for b to be generated, and
failure to generate b (f.i. due to size heuristics) is considered an
error. [ Which I guess is similar to the -o option of many other Linux
commands. ]

ISTM that the -m option could be interpreted either way. And I prefer
the interpretation of -m as: generate an external dwz file, if beneficial.

> then it
> should be an error if cc-with-tweaks doesn't manage to produce an
> executable with the proper link to this external dwz file.  Otherwise,
> the test runs with a regular non-dwzified executable, which gives a
> false sense of security about whether the feature under test works with
> dwzified executables.
> 

Agreed. Failing in cc-with-tweaks.sh in this case will list the
test-case as UNTESTED, which is a fair representation.

[ And at some point we can add support in cc-with-tweaks.sh for dwz
--devel-ignore-size (a switch only supported if dwz is compiled with
-DDEVEL) which makes it more likely to transform the input file and
generate an -m file, even if it's not beneficial, which will mean fewer
UNTESTED test-cases. ]

[ nitpick: And coming back to the false sense of security: there's no
guarantee that the DWARF that is intended to be tested by a particular
test-case is in fact transformed by dwz, even if the executable as a
whole is indeed transformed. So running the test-suite with dwz does not
guarantee that the feature under test works with dwzified executables. ]

> So this patch adds checks for that after invoking dwz.  It also removes
> the 2>&1 to allow the error message to be printed like so:
> 
>     Running /home/smarchi/src/binutils-gdb/gdb/testsuite/gdb.base/index-cache.exp ...
>     gdb compile failed, dwz: /home/smarchi/build/binutils-gdb/gdb/testsuite/outputs/gdb.base/index-cache/index-cache.dwz: .debug_info section not present
> 

Nice. I've used your patch to test, and because of this more verbose
behaviour I already noticed a new error message ("Error mmapping
multi-file temporary files"), and fixed it in dwz.

> With this patch, fewer tests will pass than before with the
> cc-with-dwz and cc-with-dwz-m boards, but those were false positives
> anyway.
> 

Ack.

> [1] Note that dwz has been patched by Tom de Vries to work correctly in
> this case, so we can use dwz master to run the test:
> 
> https://sourceware.org/git/?p=dwz.git;a=commit;h=08becc8b33453b6d013a65e7eeae57fc1881e801
> 
> gdb/ChangeLog:
> 
> 	* contrib/cc-with-tweaks.sh: Check for return value of dwz.
> ---
>  gdb/contrib/cc-with-tweaks.sh | 8 ++++++--
>  1 file changed, 6 insertions(+), 2 deletions(-)
> 
> diff --git a/gdb/contrib/cc-with-tweaks.sh b/gdb/contrib/cc-with-tweaks.sh
> index 47379cc15875..366b90918c18 100755
> --- a/gdb/contrib/cc-with-tweaks.sh
> +++ b/gdb/contrib/cc-with-tweaks.sh
> @@ -180,10 +180,14 @@ if [ "$want_index" = true ]; then
>  fi
>  
>  if [ "$want_dwz" = true ]; then
> -    $DWZ "$output_file" > /dev/null 2>&1
> +    $DWZ "$output_file" > /dev/null
> +    rc=$?
> +    [ $rc != 0 ] && exit $rc
>  elif [ "$want_multi" = true ]; then
>      cp $output_file ${output_file}.alt
> -    $DWZ -m ${output_file}.dwz "$output_file" ${output_file}.alt > /dev/null 2>&1
> +    $DWZ -m ${output_file}.dwz "$output_file" ${output_file}.alt > /dev/null
> +    rc=$?
> +    [ $rc != 0 ] && exit $rc
>      rm -f ${output_file}.alt
>  fi

I wonder if it's more reliable to test for presence of -m file rather
than exit status (and likewise, we could copy and compare for the
want_dwz case).

Also, I propose to run dwz with -q, to inhibit the countless "not
beneficial" messages. It's good to list these cases as untested, but in
the log we'd rather skip the uninteresting messages, to increase the
likelyhood that interesting messages get spotted.

Thanks,
- Tom

Patch

diff --git a/gdb/contrib/cc-with-tweaks.sh b/gdb/contrib/cc-with-tweaks.sh
index 47379cc15875..366b90918c18 100755
--- a/gdb/contrib/cc-with-tweaks.sh
+++ b/gdb/contrib/cc-with-tweaks.sh
@@ -180,10 +180,14 @@  if [ "$want_index" = true ]; then
 fi
 
 if [ "$want_dwz" = true ]; then
-    $DWZ "$output_file" > /dev/null 2>&1
+    $DWZ "$output_file" > /dev/null
+    rc=$?
+    [ $rc != 0 ] && exit $rc
 elif [ "$want_multi" = true ]; then
     cp $output_file ${output_file}.alt
-    $DWZ -m ${output_file}.dwz "$output_file" ${output_file}.alt > /dev/null 2>&1
+    $DWZ -m ${output_file}.dwz "$output_file" ${output_file}.alt > /dev/null
+    rc=$?
+    [ $rc != 0 ] && exit $rc
     rm -f ${output_file}.alt
 fi