I'm not sure that's a good way to construct documentation examples
personally, but here is an example program demonstrating that my claim
is correct, as I agree that's useful for you to have.
Feel free to edit it as you see fit.
#define _GNU_SOURCE
#define _FILE_OFFSET_BITS 64
#include <fcntl.h>
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
int main()
{
int fd = open("example_file.dat", O_WRONLY | O_CREAT | O_TRUNC, 0666);
if (fd < 0) {
printf("couldn't create example file");
exit(1);
}
int pipefd[2];
pipe(pipefd);
char ch = '\1';
write(pipefd[1], &ch, 1);
off_t offset = 0x10;
if (splice(pipefd[0], NULL, fd, &offset, 0x10, 0) < 0) {
printf("splice FAILED");
exit(1);
}
close(fd);
printf("done - new offset is %lld", (long long) offset);
}
On Mon, Aug 26, 2024 at 1:12 PM Alejandro Colomar <alx@xxxxxxxxxx> wrote:
>
> Hi Absee,
>
> On Mon, Aug 26, 2024 at 07:40:37AM GMT, Absee Seeab wrote:
> > Hi.
> > Not sure I have an idea for an example program.
> >
> > As for a paragraph documenting it, here is a patch of one suggestion,
> > though feel free to modify it or change it as needed.
>
> Maybe write a program that demonstrates that the statement is true
> without having to look at the kernel source? That will probably serve
> for the EXAMPLES section. :)
>
> Have a lovely day!
> Alex
>
> >
> > --- splice.2 2024-08-26 07:39:07.553952939 +0300
> > +++ splice.2 2024-08-26 07:38:36.503953192 +0300
> > @@ -63,7 +63,9 @@
> > .IR fd_in ;
> > in this case, the file offset of
> > .I fd_in
> > -is not changed.
> > +is not changed, and the offset pointed to by
> > +.I off_in
> > +is adjusted appropriately instead..
> > .P
> > Analogous statements apply for
> > .I fd_out
> >
> > On Sun, Aug 25, 2024 at 12:08 PM Alejandro Colomar <alx@xxxxxxxxxx> wrote:
> > >
> > > Hello Absee,
> > >
> > > On Sun, Aug 25, 2024 at 07:22:43AM GMT, Absee Seeab wrote:
> > > > Hello,
> > > >
> > > > The splice(2) manpage (at least the version at man7.org) does not
> > > > specify that off_in/off_out are written to after the operation.
> > > >
> > > > This, however, is done by the kernel (observed both by behaviour and
> > > > source code).
> > > >
> > > > Plus I see similar functions (like copy_file_range and send_file) do
> > > > document this fact, but splice does not, so it seems like a
> > > > documentation omission.
> > > >
> > > > (This caused me a bit of a debugging headache today)
> > >
> > > Would you mind adding an example program to that page, and also some
> > > paragraph that documents that?
> > >
> > > Cheers,
> > > Alex
> > >
> > > >
> > > > Thanks
> > > >
> > >
> > > --
> > > <https://www.alejandro-colomar.es/>
> >
>
> --
> <https://www.alejandro-colomar.es/>
