[oi-dev] What changed between glib versions?

[Andreas Wacknitz]

Am 31.03.22 um 17:58 schrieb Gary Mills:
> We now know which glib versions allow USB automounts and which fail
> to do so.  The question is: what has changed between these versions?
In my experience this is not fully accurate. I have USB devices
(Logitech mouse and data sticks) that work sometimes.
Funnily, another keyboard/mouse combo from Logitech works reliably with
glib-2.62.6, but another mouse (Performance MX) can even be lost without
unplugging it by
just plugging in another USB device (data stick). My two sticks
sometimes work and sometimes not. This seems to be a timing problem.

My USB devices worked with glib-2.70.0 with illumos-gate from about
three or four days ago but with the latest versions they don't seem to
work anymore.
So, this might also be a subtle timing problem.

> The bug itself is probably not in glib at all, but knowing what code
> has changed would at least put us on the right track.
>
> In the process of doing an automount, hal calls the function
> g_io_channel_read_line() within glib.  That function in turn calls the
> function g_io_channel_read_line_backend() to do the actual work.
>
> The function g_io_channel_read_line_backend() in turn calls the
> function g_io_channel_fill_buffer().
>
> Any change in any of these functions is suspect.  As far as I can tell
> by reading the code, these functions are behaving as documented.
> However, they may no longer be compatible with the operation expected
> by hal.  The compiler, or optimizer, may also be responsible.
>
>