Re: [libvirt] [PATCH] util: Extract locale-related fixes into separate functions

On Fri, Jun 23, 2017 at 10:45:30AM +0100, Daniel P. Berrange wrote: > On Fri, Jun 23, 2017 at 11:42:01AM +0200, Martin Kletzander wrote: > > On Fri, Jun 23, 2017 at 09:32:07AM +0100, Daniel P. Berrange wrote: > > > On Fri, Jun 23, 2017 at 10:21:32AM +0200, Martin Kletzander wrote: > > > > On Thu, Jun 22, 2017 at 03:49:03PM +0200, Peter Krempa wrote: > > > > > On Thu, Jun 22, 2017 at 14:36:53 +0200, Martin Kletzander wrote: > > > > > > Signed-off-by: Martin Kletzander <mkletzan(a)redhat.com&gt; > > > > > > --- > > > > > > src/util/virstring.c | 96 ++++++++++++++++++++++++++++++++-------------------- > > > > > > 1 file changed, 60 insertions(+), 36 deletions(-) > > > > > > > > > > > > diff --git a/src/util/virstring.c b/src/util/virstring.c > > > > > > index feea5be05198..6125725364f3 100644 > > > > > > --- a/src/util/virstring.c > > > > > > +++ b/src/util/virstring.c > > > > > > @@ -522,6 +522,7 @@ virStrToLong_ullp(char const *s, char **end_ptr, int base, > > > > > > #if HAVE_NEWLOCALE > > > > > > > > > > > > static locale_t virLocale; > > > > > > +static locale_t virLocaleOld; > > > > > > > > > > This is not a thread local variable ... > > > > > > > > > > > > > Oh. shoot, you're right. So we need to wrap it in our struct that will > > > > be defined conditionally as well, or allocate it on the heap, copy it > > > > there and set a pointer to it as one additional parameter or make thread > > > > local variable, but that one will need to be a pointer to the locale, so > > > > we'll need to allocate it on the heap as well. I don't like the way > > > > this is going. Anyone else feel free to continue on this if you want to > > > > clean this up, but I like the few small conditional blocks better, > > > > especially since we have them in only two functions. > > > > > > Just have virLocaleSet() return the original locale to the caller, and > > > it can pass it back in when it reverts, avoiding any global variables. > > > > > > > You cannot do that when you don't have 'locale_t', hence the > > complications. Or is the existence of locale_t somehow guaranteed by > > gnulib? If it is, then that's solved. > > Just wrap it in a typedef so we don't directly expose it in headers: > > #if HAVE_LOCALE_T > typedef virLocale locale_t > #else > typedef virLocale void > #endif > That's what I was talking about. We need this, but also a possibility to return error, so this needs to be an argument... And it's way more code that is there right now. I'm not saying NO to the change, I just that I like it like it's now (until more things need this treatment). So for others, feel free to propose changes as you like :)