| Message ID | m3zjc3akn2.fsf@sspiff.org |
|---|---|
| State | New |
| Headers | show |
On 11/07/2014 01:01 AM, Doug Evans wrote: > I was cross-testing from linux to windows7 with stdio gdbserver, > and ssh was taking longer to connect than the default remotetimeout > of 2 seconds. > > This patch adds the ability to specify remotetimeout in the board file. > > For reference sake, here's my board file. Note you can already put set GDBFLAGS "-l 15" in the board file for this. See: https://sourceware.org/gdb/wiki/TestingGDB#Running_GDB_with_a_larger_remote_serial_protocol_timeout Is there an advantage to this method? Thanks, Pedro Alves
On Fri, Nov 7, 2014 at 2:21 AM, Pedro Alves <palves@redhat.com> wrote: > On 11/07/2014 01:01 AM, Doug Evans wrote: > >> I was cross-testing from linux to windows7 with stdio gdbserver, >> and ssh was taking longer to connect than the default remotetimeout >> of 2 seconds. >> >> This patch adds the ability to specify remotetimeout in the board file. >> >> For reference sake, here's my board file. > > Note you can already put > > set GDBFLAGS "-l 15" > > in the board file for this. See: > > https://sourceware.org/gdb/wiki/TestingGDB#Running_GDB_with_a_larger_remote_serial_protocol_timeout > > Is there an advantage to this method? Yeah, it's documented in testsuite/README so I'll find it next time. 1/2 :-) I'd long since forgotten about the wiki page. Having both places being reasonably large, and neither having a reference to the other, is unfortunate. [The README mentions the wiki, but without an explicit link it's not that helpful.] I'd be happy with gutting testsuite/README and just mentioning the wiki url. Thoughts?
On 11/07/2014 03:25 PM, Doug Evans wrote: > On Fri, Nov 7, 2014 at 2:21 AM, Pedro Alves <palves@redhat.com> wrote: >> Is there an advantage to this method? > > Yeah, it's documented in testsuite/README so I'll find it next time. 1/2 :-) > :-) > I'd long since forgotten about the wiki page. > Having both places being reasonably large, and neither having a reference > to the other, is unfortunate. > [The README mentions the wiki, but without an explicit link > it's not that helpful.] > > I'd be happy with gutting testsuite/README and just mentioning the wiki url. > Thoughts? I ( still [1] ) prefer having the testsuite/README file in the tree documenting the variables that the testsuite supports, because those are tied to the specific GDB version in the tree, while the (whole) wiki doesn't really distinguish that. IMO, the wiki serves best for "recipes" and tips. It looks like testsuite/README doesn't document GDBFLAGS though. How about we fix that, and use 'set GDBFLAGS "-l 15"' as the example? I think the "Writing Tests" section of README could/should be moved to the wiki, though, e.g., to the GDBTestcaseCookbook page. [1] https://sourceware.org/ml/gdb-patches/2013-09/msg00192.html Thanks, Pedro Alves
diff --git a/gdb/testsuite/README b/gdb/testsuite/README index 9a5059a..0e4b73c 100644 --- a/gdb/testsuite/README +++ b/gdb/testsuite/README @@ -367,6 +367,10 @@ gdb,predefined_tsv The predefined trace state variables the board has. +gdb,remotetimeout + + The value for gdb's "remotetimeout" parameter, for use with gdbserver. + Testsuite Organization ********************** diff --git a/gdb/testsuite/lib/gdbserver-support.exp b/gdb/testsuite/lib/gdbserver-support.exp index 026a937..b99a594 100644 --- a/gdb/testsuite/lib/gdbserver-support.exp +++ b/gdb/testsuite/lib/gdbserver-support.exp @@ -39,6 +39,8 @@ # After GDB starts you should check global $gdbserver_gdbport for the # real port used. It is not useful if $gdbserver_reconnect_p was not set. # +# set_board_info gdb,remotetimeout +# The value for gdb's "remotetimeout" parameter. # # gdb_target_cmd @@ -47,6 +49,11 @@ proc gdb_target_cmd { targetname serialport } { global gdb_prompt + if [target_info exists gdb,remotetimeout] { + set remotetimeout [target_info gdb,remotetimeout] + gdb_test_no_output "set remotetimeout $remotetimeout" + } + set serialport_re [string_to_regexp $serialport] for {set i 1} {$i <= 3} {incr i} { send_gdb "target $targetname $serialport\n" diff --git a/gdb/testsuite/lib/mi-support.exp b/gdb/testsuite/lib/mi-support.exp index a7f7aef..0437fa5 100644 --- a/gdb/testsuite/lib/mi-support.exp +++ b/gdb/testsuite/lib/mi-support.exp @@ -335,6 +335,19 @@ proc mi_gdb_reinitialize_dir { subdir } { proc mi_gdb_target_cmd { targetname serialport } { global mi_gdb_prompt + if [target_info exists gdb,remotetimeout] { + set remotetimeout [target_info gdb,remotetimeout] + send_gdb "46-gdb-set remotetimeout $remotetimeout\n" + gdb_expect 10 { + -re ".*46-gdb-set remotetimeout $remotetimeout\r\n46\\\^done\r\n$mi_gdb_prompt$" { + verbose "Setting remotetimeout to $remotetimeout." 2 + } + timeout { + warning "Couldn't set remotetimeout to $remotetimeout." + } + } + } + set serialport_re [string_to_regexp $serialport] for {set i 1} {$i <= 3} {incr i} { send_gdb "47-target-select $targetname $serialport\n"
Hi. I was cross-testing from linux to windows7 with stdio gdbserver, and ssh was taking longer to connect than the default remotetimeout of 2 seconds. This patch adds the ability to specify remotetimeout in the board file. For reference sake, here's my board file. ---snip--- # To use this file: # bash$ cd ${build_dir}/gdb # bash$ make check RUNTESTFLAGS="--target_board=win7" # # gdbserver running over ssh. set REMOTE_HOSTNAME win7 set REMOTE_USERNAME dje set REMOTE_PORTNUM 22 set REMOTE_TMPDIR /tmp load_board_description "remote-stdio-gdbserver" # Path to the gdbserver executable. unset_board_info gdb_server_prog set_board_info gdb_server_prog "/usr/bin/gdbserver.exe" # It takes a few seconds to establish the ssh session. set_board_info gdb,remotetimeout 15 ---snip--- 2014-11-06 Doug Evans <xdje42@gmail.com> * README: Document gdb,remotetimeout. * lib/gdbserver-support.exp (gdb_target_cmd): Apply remotetimeout if specified in board description. * lib/mi-support.exp (mi_gdb_target_cmd): Ditto.