On Tue, Jan 21, 2020 at 5:40 PM Nathan Sidwell <nat...@acm.org> wrote: > > On 1/21/20 11:26 AM, Richard Biener wrote: > > On January 21, 2020 4:37:00 PM GMT+01:00, Jan Hubicka <hubi...@ucw.cz> > > wrote: > > >> I would say we want to fix that... Even GCC gets idea to put # info > >> filenames and I am sure except for HHVM there are others. > > > > Can't w just quote them somehow? > > IIUC there are two issues > > 1) Make's syntax. Its quoting is baroque, (and incomplete?). See the > comment in libcpp/mkdeps.c, which I think is curtesy of ZackW. I see > that the just-released Make 4.3 has changed something to do with # and > its need to be quoted in some contexts.
Ah, maybe it's worth to export this functionality to libiberty? > 2) Bad chars for the underlying filesystem. But, if the temp files are > on the same FS as the thing they're temping for, then there shouldn't be > a conflict. If they are different, then one might hope the temp-fs > allows a superset. Yeah, so I'd suggest to simply avoid the bad names being created from the WPA stage and instead sanitize it there already. That would make it also consistent when debugging. Not to say that using a tempfile by default when not doing -save-temps would be appropriate anyway. Richard. > nathan > > -- > Nathan Sidwell
