path: root/open_issues/gdb.mdwn

[[!meta copyright="Copyright © 2007, 2008, 2010, 2011, 2012, 2013 Free Software
Foundation, Inc."]]

[[!meta license="""[[!toggle id="license" text="GFDL 1.2+"]][[!toggleable
id="license" text="Permission is granted to copy, distribute and/or modify this
document under the terms of the GNU Free Documentation License, Version 1.2 or
any later version published by the Free Software Foundation; with no Invariant
Sections, no Front-Cover Texts, and no Back-Cover Texts.  A copy of the license
is included in the section entitled [[GNU Free Documentation
License|/fdl]]."]]"""]]

[[!tag open_issue_gdb]]

Here's what's to be done for maintaining GNU GDB.

[[!toc levels=2]]


# [[General information|/gdb]]


# [[Sources|source_repositories/gdb]]


# Configuration

<!--

git checkout reviewed
git log --reverse --topo-order --pretty=fuller --stat=$COLUMNS,$COLUMNS -w -p -C --cc ..sourceware/master
-i
/^commit |^merge:|^---$|hurd|linux|nacl|nptl|glibc|gs:|gnu-nat|i386gnu

-->

Last reviewed up to the [[Git mirror's 6b25dae901ddedb3f330803542d3eac73cdcae4b
(2013-03-13) sources|source_repositories/gdb]].

  * Globally

      * a.out, COFF, PE image support and 64 bit support are not interesting.

      * In the testsuites, `.exp` and `.d` files very likely should not only
        care for `*-*-linux*`, but also `*-*-gnu*`.  (If the need to be
        conditionalized like this at all.)

  * `bfd/`

    See [[binutils]].

  * `libdecnumber/`

    Should/can probably align to GNU/Linux.

  * Have a look at config/i386/i386gnu.mh.

  * configure.tgt

      * glibc-tdep et al. also for GNU/Hurd?

  * [[gdbserver]]

  * 82763a3d329b0d342d0273941b1521be9ef0c604 »MODIFIED is unknown, pass it as
    true.«

  * Configure so that Debian system's `/usr/lib/debug/[...]` will be loaded
    automatically.


# Build

Here's a log of a GDB build run; this is from our [[Git
repository|source_repositories/gdb]]'s `tschwinge/Ferry_Tagscherer` branch,
commit 6b25dae901ddedb3f330803542d3eac73cdcae4b (2013-03-13), run on
kepler.SCHWINGE and coulomb.SCHWINGE.

    $ export LC_ALL=C
    $ ../Ferry_Tagscherer/configure --prefix="$PWD".install SHELL=/bin/dash CC=gcc-4.7 CXX=g++-4.7 --disable-werror 2>&1 | tee log_build
    [...]
    $ make 2>&1 | tee log_build_
    [...]

Different hosts may default to different shells and compiler versions; thus
harmonized.

There are several occurences of *error: dereferencing type-punned pointer will
break strict-aliasing rules* in the MIG-generated stub files; thus no `-Werror`
until that is resolved ([[strict_aliasing]]).

This takes up around 210 MiB and needs roughly 9 min on kepler.SCHWINGE and 25
min on coulomb.SCHWINGE.

<!--

    $ (make && touch .go-install) 2>&1 | tee log_build_ && test -f .go-install && (make install && touch .go-test) 2>&1 | tee log_install && test -f .go-test && make -k check 2>&1 | tee log_test

-->


## Analysis

x86 GNU/Linux' and GNU/Hurd's configurations are slightly different, thus mask
out most of the differences that are due to GNU/Linux supporting more core file
formats and more emulation vectors.

    $ toolchain/logs/process gdb build

  * Why do we specify `-D_GNU_SOURCE`, and GNU/Linux doesn't?

  * GNU/Linux: `gdb/symfile-mem.c` for vDSO.

  * GNU/Linux: `gdb/i386-nat.c` for hardware breakpoints, etc. -- we should
    probably use that, too.  Related to Samuel's Hurd GDB patch?

  * `gdb/gnu-nat.c`

        gnu-nat.c: In function 'proc_set_exception_port':
        gnu-nat.c:409:3: warning: format '%d' expects argument of type 'int', but argument 8 has type 'mach_port_t' [-Wformat]
        gnu-nat.c: In function 'proc_steal_exc_port':
        gnu-nat.c:449:7: warning: format '%d' expects argument of type 'int', but argument 8 has type 'mach_port_t' [-Wformat]
        gnu-nat.c:470:7: warning: format '%d' expects argument of type 'int', but argument 8 has type 'mach_port_t' [-Wformat]
        gnu-nat.c: In function 'make_proc':
        gnu-nat.c:583:7: warning: format '%d' expects argument of type 'int', but argument 2 has type 'mach_port_t' [-Wformat]
        gnu-nat.c:586:7: warning: format '%d' expects argument of type 'int', but argument 8 has type 'mach_port_t' [-Wformat]
        gnu-nat.c: In function 'inf_set_pid':
        gnu-nat.c:761:3: warning: format '%d' expects argument of type 'int', but argument 7 has type 'task_t' [-Wformat]
        gnu-nat.c: In function 'inf_validate_procs':
        gnu-nat.c:1085:6: warning: format '%d' expects argument of type 'int', but argument 8 has type 'thread_t' [-Wformat]
        gnu-nat.c: In function 'inf_signal':
        gnu-nat.c:1349:4: warning: format '%d' expects argument of type 'int', but argument 7 has type 'thread_t' [-Wformat]
        gnu-nat.c:1349:4: warning: format '%d' expects argument of type 'int', but argument 8 has type 'thread_t' [-Wformat]
        gnu-nat.c: In function 'S_exception_raise_request':
        gnu-nat.c:1668:3: warning: format '%d' expects argument of type 'int', but argument 7 has type 'thread_t' [-Wformat]
        gnu-nat.c:1668:3: warning: format '%d' expects argument of type 'int', but argument 8 has type 'task_t' [-Wformat]
        gnu-nat.c:1705:8: warning: format '%d' expects argument of type 'int', but argument 7 has type 'mach_port_t' [-Wformat]
        gnu-nat.c:1711:8: warning: format '%d' expects argument of type 'int', but argument 7 has type 'mach_port_t' [-Wformat]
        gnu-nat.c: In function 'do_mach_notify_dead_name':
        gnu-nat.c:1762:3: warning: format '%d' expects argument of type 'int', but argument 7 has type 'mach_port_t' [-Wformat]
        gnu-nat.c: In function 'gnu_write_inferior':
        gnu-nat.c:2383:8: warning: format '%x' expects argument of type 'unsigned int', but argument 2 has type 'vm_address_t' [-Wformat]
        gnu-nat.c:2393:8: warning: format '%x' expects argument of type 'unsigned int', but argument 2 has type 'vm_address_t' [-Wformat]
        gnu-nat.c: In function 'steal_exc_port':
        gnu-nat.c:2864:5: warning: format '%d' expects argument of type 'int', but argument 2 has type 'mach_port_t' [-Wformat]

    
  * fe19822761b4635f392875a186e48af446b40f41..7a63e9515491f21eaf07301df87d389def20e317:

    `-Wmissing-prototypes`

        gnu-nat.c: At top level:
        gnu-nat.c:643:1: warning: no previous prototype for 'make_inf' []
        gnu-nat.c: At top level:
        gnu-nat.c:879:1: warning: no previous prototype for 'inf_set_traced' []
        gnu-nat.c:980:1: warning: no previous prototype for 'inf_port_to_thread' []
        gnu-nat.c: At top level:
        gnu-nat.c:1748:1: warning: no previous prototype for 'inf_task_died_status' []
        gnu-nat.c: At top level:
        gnu-nat.c:2273:1: warning: no previous prototype for 'gnu_read_inferior' []
        gnu-nat.c:2319:1: warning: no previous prototype for 'gnu_write_inferior' []
        gnu-nat.c: At top level:
        gnu-nat.c:3415:1: warning: no previous prototype for '_initialize_gnu_nat' []
        notify_S.c:305:24: warning: no previous prototype for 'notify_server' []
        notify_S.c:341:28: warning: no previous prototype for 'notify_server_routine' []
        process_reply_S.c:343:24: warning: no previous prototype for 'process_reply_server' []
        process_reply_S.c:379:28: warning: no previous prototype for 'process_reply_server_routine' []
        msg_reply_S.c:165:24: warning: no previous prototype for 'msg_reply_server' []
        msg_reply_S.c:201:28: warning: no previous prototype for 'msg_reply_server_routine' []
        exc_request_S.c:157:24: warning: no previous prototype for 'exc_server' []
        exc_request_S.c:193:28: warning: no previous prototype for 'exc_server_routine' []

  * `dlopen`/`-ldl`

        -checking for library containing dlopen... none required
        +checking for library containing dlopen... -ldl

  * `O_NOFOLLOW`

    First seen in
    20f498edfd7e57d3297febcf9c7c7d667cc74239..69a5e2b022c7d15ec4c7c49e6f53a8d924d3b72b:

        -checking for working fcntl.h... yes
        +checking for working fcntl.h... no (bad O_NOFOLLOW)

    [[!taglink open_issue_glibc]]?

  * Why does GNU/Linux have an additional `-ldl -rdynamic` when linking `gdb`?

  * 00e1c4298d2b6fe040a9a970e98349602b12ddbf..6b25dae901ddedb3f330803542d3eac73cdcae4b:

        +../../Ferry_Tagscherer/gdb/gnu-nat.c: In function 'info_port_rights':
        +../../Ferry_Tagscherer/gdb/gnu-nat.c:3058:11: warning: passing argument 1 of 'parse_to_comma_and_eval' from incompatible pointer type [enabled by default]
        +In file included from ../../Ferry_Tagscherer/gdb/breakpoint.h:23:0,
        +                 from ../../Ferry_Tagscherer/gdb/inferior.h:37,
        +                 from ../../Ferry_Tagscherer/gdb/gnu-nat.c:56:
        +../../Ferry_Tagscherer/gdb/value.h:729:22: note: expected 'const char **' but argument is of type 'char **'


# Install

    $ make install 2>&1 | tee log_install
    [...]

This takes up around 50 MiB, and needs roughly 1 min on kepler.SCHWINGE and 2
min on coulomb.SCHWINGE.


## Analysis

    $ toolchain/logs/process gdb install

  * `libtool: finish`: `ldconfig` is not run for the Hurd.


# Testsuite

    $ make -k check 2>&1 | tee log_test
    [...]

This needs roughly 14 min on kepler.SCHWINGE and 70 min on coulomb.SCHWINGE.

When running `make -k check 2>&1 | tee log_test`, at the end of the testsuite
the `tee` process does not terminate if there are still stray leftover
processes that [have their stdout/stderr
open](http://sourceware.org/ml/gdb-patches/2012-10/msg00489.html).  `kill`ing
these (`SIGKILL` may be needed), makes the `tee` process terminate, too.  On
GNU/Hurd, these generally are `gdb.multi/watchpoint-multi`, and an unknown
(`?`) one ("57 PIDs before" `expect [...] gdb.cp`).


## Analysis

    $ toolchain/logs/process gdb test

  * `gdb.base/attach-pie-misread.exp`

    Is only run for GNU/Linux; needs [[prelink]].

  * Disabled

      * `gdb.base/readline.exp`

        [[term_blocking]] issue.

      * `gdb.base/sigall.exp`

        From `send signal TSTP` on, all FAIL running into timeouts.

      * `gdb.python/py-inferior.exp` (mostly disabled)

            Running ../../../Ferry_Tagscherer/gdb/testsuite/gdb.python/py-inferior.exp ...
            [...]
            python print 'result =', i0.was_attached
            result = False
            (gdb) PASS: gdb.python/py-inferior.exp: test Inferior.was_attached
            python print i0.threads ()
            (<gdb.InferiorThread object at 0x61170>, <gdb.InferiorThread object at 0x61160>)
            (gdb) FAIL: gdb.python/py-inferior.exp: test Inferior.threads
            break check_threads
            Breakpoint 2 at 0x8048869: file ../../../Ferry_Tagscherer/gdb/testsuite/gdb.python/py-inferior.c, line 61.
            (gdb) continue
            Continuing.
            [New Thread 25670.6]
            [New Thread 25670.7]
            [New Thread 25670.8]
            [New Thread 25670.9]
            [New Thread 25670.10]
            [New Thread 25670.11]
            [New Thread 25670.12]
            [New Thread 25670.13]
            
            Breakpoint 2, check_threads (barrier=0x15ff144) at ../../../Ferry_Tagscherer/gdb/testsuite/gdb.python/py-inferior.c:61
            61        pthread_barrier_wait (barrier);
            (gdb) PASS: gdb.python/py-inferior.exp: continue to breakpoint: cont to check_threads
            python print len (i0.threads ())
            10
            (gdb) FAIL: gdb.python/py-inferior.exp: test Inferior.threads 2
            break 28
            Breakpoint 3 at 0x80487c2: file ../../../Ferry_Tagscherer/gdb/testsuite/gdb.python/py-inferior.c, line 28.
            (gdb) continue
            Continuing.
            FAIL: gdb.python/py-inferior.exp: continue to breakpoint: cont to Break here. (timeout)
            python addr = gdb.selected_frame ().read_var ('str')
            FAIL: gdb.python/py-inferior.exp: read str address (timeout)
            [All following tests FAIL with timeout.]
            FAIL: gdb.python/py-inferior.exp: Switch to first inferior (timeout)
            remove-inferiors 3
            FAIL: gdb.python/py-inferior.exp: Remove second inferior (timeout)

        At this point, the system hangs; no new processes can be spawned, so
        perhaps an issue with the exec server.

  * Linux syscall usage, `<asm/unistd.h>`

      * `UNSUPPORTED: gdb.threads/ia64-sigill.exp: Couldn't compile ../../../master/gdb/testsuite/gdb.threads/ia64-sigill.c: unrecognized error`

      * `UNSUPPORTED: gdb.threads/siginfo-threads.exp: Couldn't compile ../../../Ferry_Tagscherer/gdb/testsuite/gdb.threads/siginfo-threads.c: unrecognized error`

      * `gdb.threads/sigstep-threads.c`

        Also uses `tgkill`.

      * `UNSUPPORTED: gdb.threads/watchpoint-fork.exp: parent: multithreaded: Couldn't compile ../../../Ferry_Tagscherer/gdb/testsuite/gdb.threads/watchpoint-fork-mt.c ../../../Ferry_Tagscherer/gdb/testsuite/gdb.threads/watchpoint-fork-parent.c: unrecognized error`

  * `UNSUPPORTED: gdb.threads/multi-create.exp: Couldn't compile ../../../master/gdb/testsuite/gdb.threads/multi-create.c: unrecognized error`
        ../../../master/gdb/testsuite/gdb.threads/multi-create.c: In function 'create_function':
        ../../../master/gdb/testsuite/gdb.threads/multi-create.c:46:39: error: 'PTHREAD_STACK_MIN' undeclared (first use in this function)
        ../../../master/gdb/testsuite/gdb.threads/multi-create.c:46:39: note: each undeclared identifier is reported only once for each function it appears in
        ../../../master/gdb/testsuite/gdb.threads/multi-create.c: In function 'main':
        ../../../master/gdb/testsuite/gdb.threads/multi-create.c:73:39: error: 'PTHREAD_STACK_MIN' undeclared (first use in this function)

  * `UNSUPPORTED: gdb.threads/staticthreads.exp: Couldn't compile ../../../master/gdb/testsuite/gdb.threads/staticthreads.c: unrecognized error`

        ../../../master/gdb/testsuite/gdb.threads/staticthreads.c: In function 'main':
        ../../../master/gdb/testsuite/gdb.threads/staticthreads.c:52:37: error: 'PTHREAD_STACK_MIN' undeclared (first use in this function)
        ../../../master/gdb/testsuite/gdb.threads/staticthreads.c:52:37: note: each undeclared identifier is reported only once for each function it appears in

  * `UNSUPPORTED: gdb.threads/create-fail.exp: Couldn't compile ../../../Ferry_Tagscherer/gdb/testsuite/gdb.threads/create-fail.c: unrecognized error`

        [...]/gdb.threads/create-fail.c:77: undefined reference to `pthread_attr_setaffinity_np'
        [...]/gdb.threads/create-fail.c:83: undefined reference to `pthread_create'

  * `UNTESTED: gdb.base/longest-types.exp: longest-types.exp`

        ../../../Ferry_Tagscherer/gdb/testsuite/gdb.base/longest-types.c:20:8: error: size of array 'buf' is too large

    Also on GNU/Linux.

  * `FAIL: gdb.base/jit.exp: PIE: one_jit_test-1: Can't run to main`

        (gdb) break main
        Breakpoint 1 at 0xb84: file ../../../Ferry_Tagscherer/gdb/testsuite/gdb.base/jit-main.c, line 128.
        (gdb) run 
        Starting program: /media/erich/home/thomas/tmp/gdb/tschwinge/Ferry_Tagscherer.build/gdb/testsuite/gdb.base/jit-main 
        Cannot access memory at address 0x393
        Cannot access memory at address 0x38f
        (gdb) FAIL: gdb.base/jit.exp: PIE: one_jit_test-1: Can't run to main

    [[GCC/PIE]].

    Is the following supposed to terminate in this way?

        (gdb) break main
        Breakpoint 1 at 0x675: file ../../../Ferry_Tagscherer/gdb/testsuite/gdb.base/attach-pie-noexec.c, line 23.
        (gdb) run 
        Starting program: /media/erich/home/thomas/tmp/gdb/tschwinge/Ferry_Tagscherer.build/gdb/testsuite/gdb.base/attach-pie-noexec 
        Cannot access memory at address 0x6c626172
        Cannot access memory at address 0x6c62616e
        (gdb) testcase ../../../Ferry_Tagscherer/gdb/testsuite/gdb.base/attach-pie-noexec.exp completed in 3 seconds

  * `solib-event stop`

        Running ../../../Ferry_Tagscherer/gdb/testsuite/gdb.mi/mi-catch-load.exp ...
        PASS: gdb.mi/mi-catch-load.exp: breakpoint at main
        PASS: gdb.mi/mi-catch-load.exp: mi runto main
        PASS: gdb.mi/mi-catch-load.exp: catch-load: auto-solib-add on
        PASS: gdb.mi/mi-catch-load.exp: catch-load: catch load
        FAIL: gdb.mi/mi-catch-load.exp: catch-load: solib-event stop
        PASS: gdb.mi/mi-catch-load.exp: breakpoint at main
        PASS: gdb.mi/mi-catch-load.exp: mi runto main
        PASS: gdb.mi/mi-catch-load.exp: catch-unload: auto-solib-add on
        PASS: gdb.mi/mi-catch-load.exp: catch-unload: catch unload
        FAIL: gdb.mi/mi-catch-load.exp: catch-unload: solib-event stop

        *stopped,reason="signal-received",signal-name="SIGSEGV",signal-meaning="Segmentation fault",frame={addr="0x00014add",func="??",args=[],from="/lib/ld.so"},thread-id="4",stopped-threads="all"

TODO.