Re: [Xen-devel] [PATCH v2-resend 11/30] libxl: ocaml: allocate a long lived libxl context.

[Rob Hoes <Rob.Hoes@xxxxxxxxxx>]

> Rob Hoes writes ("[Xen-devel] [PATCH v2-resend 11/30] libxl: ocaml:
> allocate a long lived libxl context."):
> > Rather than allocating a new context for every libxl call begin to
> > switch to a model where a context is allocated by the caller and may
> > then be used for multiple calls down into the library.
> ...
> > @@ -59,6 +65,8 @@ static void log_destroy(struct xentoollog_logger
> *logger)
> >     lg.logger.vmessage = log_vmessage; \
> >     lg.logger.destroy = log_destroy; \
> >     lg.logger.progress = NULL; \
> > +   lg.log_offset = 0; \
> > +   memset(&lg.log_buf,0,sizeof(lg.log_buf));       \
> 
> Is this in the wrong patch ?
> 
> @@ -77,7 +85,7 @@ static char * dup_String_val(caml_gc *gc, value s)
> >     c = calloc(len + 1, sizeof(char));
> >     if (!c)
> >             caml_raise_out_of_memory();
> > -   gc->ptrs[gc->offset++] = c;
> > +   if (gc) gc->ptrs[gc->offset++] = c;
> 
> I don't understand this at all.  Is it going to become local to call this 
> without a
> gc and if so what is its function in that case ?
> 
> > @@ -94,9 +102,41 @@ static void failwith_xl(char *fname, struct
> > caml_logger *lg)  {
> >     char *s;
> >     s = (lg) ? lg->log_buf : fname;
> > +   printf("Error: %s\n", fname);
> >     caml_raise_with_string(*caml_named_value("xl.error"), s);  }
> 
> I don't understand why this hunk is in this patch, either.

None of these hunks should be there. Both the logger and gc bits are removed 
entirely in the following patch, and that printf is a debugging line I forgot 
to remove.

I probably made some mistakes while reorganising and squashing some patches. 
I'll remove those hunks.

Cheers,
Rob


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel