On Thu, Mar 31, 2022 at 1:44 PM Mark Wielaard <m...@klomp.org> wrote:
> Just a question about this part:
>
> > + /* If Content-Length is -1, try to get the size from
> > + X-Debuginfod-Size */
> > + if (dl_size == -1 && c->winning_headers != NULL)
> > + {
> > + double xdl;
> > + char *hdr = strstr(c->winning_headers, "x-debuginfod-size");
> > +
> > + if (hdr != NULL
> > + && sscanf(hdr, "x-debuginfod-size: %lf", &xdl) == 1)
> > + dl_size = (xdl >= (double)(LONG_MAX+1UL) ? LONG_MAX :
> > (long)xdl);
> > + }
> > + }
>
> In debuginfod.cxx the header is spelled all uppercase as "X-DEBUGINFOD-
> SIZE" which is also what is checked for in the run-debuginfod-response-
> headers.sh test. So shouldn't the above also be all uppercase or should
> you use strcasestr?
strcasestr is definitely better here. I meant to replace strstr with it
but it looks like I missed that.
> When using sscanf why are you using a double and %lf? Isn't it simpler
> to use a long and %ld?
It was done out of an (excessive) abundance of caution in case the
value from the header was greater than LONG_MAX. x-debuginfod-size is
derived from an off_t value so this currently isn't possible, so yes it
would be simpler to stick with long and %ld. Fixed.
> Is there a way to test this easily? When would Content-Length not be
> available?
AFAIK Content-Length isn't available only when a debuginfod server is
configured with an httpd proxy that compresses files on-the-fly. If the
response headers are finalized before compression is finished,
Content-Length won't be present.
Short of setting up an httpd-proxied server in the testsuite I'm not sure
exactly how to test this. We currently have tests that check for the
presence of x-debuginfod-size and we could add tests to at least verify
that the value in the header matches the uncompressed size of the file
being transferred.
Aaron
