On 3/11/19 10:05 AM, Jakub Jelinek wrote: > On Mon, Mar 11, 2019 at 09:57:36AM +0100, Martin Liška wrote: >> On 3/11/19 9:30 AM, Jakub Jelinek wrote: >>> On Mon, Mar 11, 2019 at 09:13:57AM +0100, Martin Liška wrote: >>>> The patch adds a lot of option name wrapping in string format messages. I >>>> added a new contrib >>>> script (contrib/check-internal-format-escaping.py) that is parsing gcc.pot >>>> file and reports >>>> errors. >>>> >>>> Patch can bootstrap on x86_64-linux-gnu and survives regression tests. >>>> Apart from that I built all cross compilers and compared all warnings so >>>> that >>>> I don't introduce a bootstrap error. It's expected that various >>>> target-specific >>>> tests will need wrapping in scanned patterns. >>>> >>>> Is it fine for next stage1? >>> >>> Generally looks good to me, but I'm not sure about corner cases like: >>> %<-misr-secure=X%>, shouldn't the X be after %>? X is not what users would >>> type. Or reword these to %<-misr-secure=%s%> argument not in between 0 and >>> 23 or similar. >> >> Well, in order to make it consistent, I would put the closing '%>' after >> the whole option=argument expression. > > The problem is that the X is not what people should write literally on the > command line, unlike everything else we put in between the quotes. That is > why I suggest to rework it like other targets do, where they actually print > the argument the user specified (which should be in between quotes) and > don't use any X in the wording. > > Jakub >
Now I understand that, thanks. Sending updated patch. Martin
0001-Wrap-option-names-in-gcc-internal-messages-with-and.patch.bz2
Description: application/bzip
