[dpdk-dev] [PATCH] checkpatch: re-enable warnings about split long strings

[Bruce Richardson]

On Mon, Oct 02, 2017 at 06:21:06PM +0200, Adrien Mazarguil wrote:
> On Mon, Oct 02, 2017 at 02:46:24PM +0100, Bruce Richardson wrote:
> > On Mon, Oct 02, 2017 at 01:53:17PM +0200, Adrien Mazarguil wrote:
> > > Hi Stephen,
> > > 
> > > On Fri, Sep 29, 2017 at 08:37:49AM -0700, Stephen Hemminger wrote:
> > > > The Linux kernel style policy about strings is that strings should
> > > > be always put on one line. This makes sense since a typical use case
> > > > is for a user to type the error message into a search engine or
> > > > grep, and it won't be found if split across lines.  This patch just
> > > > re-enables that check.
> > > > 
> > > > Yes, lots of DPDK code now splits strings, that doesn't make it
> > > > right.
> > > > 
> > > > Signed-off-by: Stephen Hemminger <sthemmin at microsoft.com> ---
> > > > devtools/checkpatches.sh | 1 - 1 file changed, 1 deletion(-)
> > > > 
> > > > diff --git a/devtools/checkpatches.sh b/devtools/checkpatches.sh
> > > > index a56c41a301c0..3e6081dd673e 100755 ---
> > > > a/devtools/checkpatches.sh +++ b/devtools/checkpatches.sh @@ -44,7
> > > > +44,6 @@ options="$options --show-types" options="$options
> > > > --ignore=LINUX_VERSION_CODE,FILE_PATH_CHANGES,\
> > > > VOLATILE,PREFER_PACKED,PREFER_ALIGNED,PREFER_PRINTF,\
> > > > PREFER_KERNEL_TYPES,BIT_MACRO,CONST_STRUCT,\
> > > > -SPLIT_STRING,LONG_LINE_STRING,\
> > > > LINE_SPACING,PARENTHESIS_ALIGNMENT,NETWORKING_BLOCK_COMMENT_STYLE,\
> > > > NEW_TYPEDEFS,COMPARISON_TO_NULL"
> > > 
> > > I'm not sure, given that the main reason for splitting strings in the
> > > first place is to avoid LONG_LINE_STRING warnings, I think we must
> > > choose between the two options. If split strings are not allowed, then
> > > long lines must be.
> > > 
> > > Since checkpatches.sh is used by various automated scripts to complain
> > > loudly about problems in submissions, the above change prevents
> > > maintainers from writing long string at all (can't split and can't go
> > > past 80 columns).
> > > 
> > > As a result, they will be tempted to cripple their code with nasty
> > > workarounds to shut up checkpatches.sh, we don't want that to happen.
> > > 
> > > Also I think the reasons stated by original commit cf75514c8e2e are
> > > still relevant. My vote would be to keep things as is.
> > > 
> > In my experience, checkpatch is smart enough to recognise when a long
> > line overflows the 80 character limit because of a single long string,
> > so the two options are not mutually exclusive. In other words, long
> > lines are not allowed except in the case where shortening the line
> > involves splitting a string. There may be a small amount of work in
> > getting checkpatch happy, i.e. by putting the string on a line on it's
> > own, but we can indeed have our cake and eat it too in this case.
> 
> I can't seem to get around warnings without ignoring either SPLIT_STRING or
> LONG_LINE_STRING as of Linux v4.14-rc3's checkpatch.pl. I think you can only
> get around them by fooling it somehow. You really need to ignore at least
> LONG_LINE_STRING to meet the requirements of the commit log.
> 
> However SPLIT_STRING still looks necessary to address part of cf75514c8e2e
> ("devtools: ignore warning on long log string"):
> 
>  "...lines that make use of PRIx64 with string concatenation will still be
>   flagged if the beginning of the last string fragment begins after the 80
>   character threshold."
> 
> It's not all that uncommon in my opinion.
> 
If you have PRIx64 in it, it's not part of a literal string you would
grep, so it's reasonable to split there. The user cannot know what the
specific %x formatting character used is.

/Bruce