Re: [PATCH 0/2] Better handling of slow remote transfers

[Pedro Alves <palves at redhat dot com>]

On 08/11/2015 06:22 PM, Doug Evans wrote:
> Gary Benson writes:
>   > Hi all,
>   >
>   > Since March or so GDB has been able to access inferior binaries for
>   > remote targets without having to be explicitly told to.  This caused
>   > problems for some people with slow connections:
>   >
>   >   https://sourceware.org/ml/gdb/2015-07/msg00038.html
>   >
>   > The first patch in this series adds the warning messages requested
>   > in that thread.  The second commit should make long transfers
>   > interruptible.
>   >
>   > Built and regtested on RHEL 6.6 x86_64.
>   >
>   > Ok to commit?
> 
> For 7.10, one thought is to maintain the behaviour of 7.9
> and give ourselves more time to address this.

Agreed.  My opinion, as expressed elsewhere in the gdb@ thread,
is that in order to be able to have target: by default in 7.10, we'd
try to both sort out the interruptibility and add a suggestive
warning, assuming both were easy to do, and not invasive,
and check if that would be a sufficient resolution.  Seems like
the interruptibility issue isn't trivial to solve, so I think we
need to go do that -- change the default sysroot back to "", (and tweak
the docs/NEWS accordingly), and buy some time to sort this out on master.
Users can still then put "set sysroot target:" in .gdbinit with 7.10,
integrators should still be able to build with --with-sysroot=target: (or
revert the future-default-sysroot-reversion patch), and we can
continue addressing identified issues until "target:" (or something
around it, maybe building up on Jan's buildid work) can be made
the default, on master.

> IOW, can we have (or is there already) a configure
> option that controls the default behaviour,
> and can we default it to what 7.9 does
> (not auto-fetch files) ?

I think that would be the existing --with-sysroot.

Thanks,
Pedro Alves