On 2023-09-08, Sebastien Marie wrote:
> "rpath" is a bit odd in xeyes(1) normal behaviour (but it will be required on
> X11 error, as if I remember well, error codes are "translated" to message by
> reading some file).
Exactly. X11 reads the error database `/usr/X11R6/share/X11/XErrorDB`
if it needs to issue an error.
However, we can force xeyes(1) to read the error database into memory at
initialization before pledging; so if an error occurs, it would not need
to read the database again.
Quoting from `xenocara/app/xclock/xclock.c`:
> {
> /* force reading of XErrorDB into memory to avoid adding "rpath" to
> pledge below */
> char buf[1];
>
> (void)XGetErrorDatabaseText(XtDisplay(toplevel), "XProtoError", "0", "",
> buf, 1);
> }
> if (pledge("stdio", NULL) == -1)
> err(1, "pledge");
XGetErrorDatabaseText uses a static variable to hold the database.
This variable is initialized in the first call, and then used by
the following calls.
On 2023-09-08, Sebastien Marie wrote:
> For me, you are pledging too early (before initialization). It should be done
> at
> least after calling XtAppInitialize(3).
>
> It will be the main limitation for a tool like `abstain`. pledge(2) should be
> called *after* initialization, and not at the beginning of the program.
I also think that.
If we pledge(2) after initializing the connection to the X Server, we
can drop the "unix" and "inet" promises.
Lucas de Sena
