This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
[Bug bootstrap/16253] bootstrapping java does not work when using GNU binutils
- From: "geos at epost dot de" <gcc-bugzilla at gcc dot gnu dot org>
- To: gcc-bugs at gcc dot gnu dot org
- Date: 19 Jul 2004 17:58:32 -0000
- Subject: [Bug bootstrap/16253] bootstrapping java does not work when using GNU binutils
- References: <20040628212020.16253.geos@epost.de>
- Reply-to: gcc-bugzilla at gcc dot gnu dot org
------- Additional Comments From geos at epost dot de 2004-07-19 17:58 -------
Subject: Re: bootstrapping java does not work when using
GNU binutils
> Hmm, so ncargs must just be an IRIX 6 thing. Is there some other way
> of controlling the command-line limit?
maybe; I would guess this is a shell feature?
Anyway, I can browse google for it; I never needed that before.
>
> As you'll probably have guessed, I don't have access to an IRIX 5 box
> myself, so can you try to debug it a bit more at your end? E.g., does
> it fail if you cut-&-paste the failing command line into a shell?
OK, I will try it on the next weekend, when I have enough time.
On a 50 MHZ R4k it takes a few hours for one run.
>
> FWIW, there are successful build reports with binutils:
>
> http://gcc.gnu.org/ml/gcc-testresults/2004-07/msg00761.html
>
> Given that the failure is in a long, but otherwise innocuous, shell command,
> and that that command would be shorter when not using binutils, it really
> does look like a limits issue to me. One possible workaround would be to
> move the gcc sources so that the full pathname to them is shorter.
I will try that, using e.g. /tmp, but it will take time.
Anyway, I'll come back to you when I need further help or have further input.
BTW, /usr/include/limits.h says:
/*
* ARG_MAX
* CHILD_MAX
* OPEN_MAX
* STREAM_MAX
*
* PATH_MAX
* NAME_MAX
* LINK_MAX
* - are available only thru sysconf/pathconf
*/
...
#ifndef PATH_MAX
#define PATH_MAX 1024 /* max # of characters in a path name */
#endif
#ifndef ARG_MAX
#define ARG_MAX 5120
#endif
the limits man page says:
#define ARG_MAX 10240 /* max length of arguments to exec */
(it seems the man page is wrong by a factor of 2 here; or maybe the 5120
above is only the last resort value. Is ARG_MAX of relevance to the gcc
problem? Is that what you are looking for?
lorenz 28% sysconf
ARG_MAX 20480
CHILD_MAX 150
CLK_TCK 100
NGROUPS_MAX 16
OPEN_MAX 200
JOB_CONTROL 1
SAVED_IDS 1
VERSION 199009
ASYNCHRONOUS_IO 1
AIO_LISTIO_MAX 255
AIO_MAX 4
REALTIME_SIGNALS 1
RTSIG_MAX 16
SIGQUEUE_MAX 32
LOGNAME_MAX 8
PASS_MAX 8
TZNAME_MAX 8
NACLS_MAX 0
PAGESIZE 4096
NPROC_CONF 1
NPROC_ONLN 1
STREAM_MAX 200
XOPEN_VERSION 3
ACL 0
AUDIT 0
INF 0
MAC 0
CAP 0
IP_SECOPTS 0
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=16253