Fwd: Re: [PATCH] Use binary mode to read .log/.sum

[Serhei Makarov]

----- Original message -----
From: Serhei Makarov <me@serhei.io>
To: Keith Seitz <keiths@redhat.com>
Subject: Re: [PATCH] Use binary mode to read .log/.sum
Date: Wednesday, September 23, 2020 8:59 PM

On Wed, Sep 23, 2020, at 5:06 PM, Keith Seitz via Bunsen wrote:
> diff --git a/scripts-master/gdb/parse_dejagnu.py 
> b/scripts-master/gdb/parse_dejagnu.py
> index 5218e9a..c9f30ac 100755
> --- a/scripts-master/gdb/parse_dejagnu.py
> +++ b/scripts-master/gdb/parse_dejagnu.py
> @@ -65,11 +65,14 @@ def get_outcome_line(testcase):
>  datestamp_format = '%a %b %d %H:%M:%S %Y'
>  
>  def openfile_or_xz(path):
> +    # Read in bary mode to suppress encoding problems that might occur
> +    # from reading gdb.{log,sum}. Sometimes inferiors or gdb can just output
> +    # garbage bytes.
>      if os.path.isfile(path):
> -        return open(path, mode='rt')
> +        return open(path, mode='rb')
>      elif os.path.isfile(path+'.xz'):
> -        return lzma.open(path+'.xz', mode='rt')
> -    return open(path, mode='rt') # XXX trigger default error
> +        return lzma.open(path+'.xz', mode='rb')
> +    return open(path, mode='rb') # XXX trigger default error
Patch LGTM.

In retrospect, the need to deal with encodings is a major annoyance about I/O in recent Python versions.
I have had to make prior fixes for similar issues, and I'm unsurprised that there is still more to fix.