problems with nALFS

Reinhard bookreader at gmx.com
Fri Apr 9 23:03:50 PDT 2004


On Saturday 10 April 2004 03:50, Kevin P. Fleming wrote:
> Reinhard wrote:
> > Using nALFS it fails at binutils-pass1 with signal 11 during configure.
>
> If you are using nALFS CVS HEAD, these problems are known and will be
> fixed in the next few days (probably Sunday evening). If not, then this
> is a serious problem in nALFS 1.2.2.

Sorry, I probably missed that mails.
Yes I'm using CVS-HEAD.

Thanks for the hint. I was a bit wired about the problems and searching for 
problems on the host-side.

> > Also, I'm not that used to linux-dev-tools, that I would know, how to get
> > the trap of a child-process. Running nALFS in gdb it shows me the code,
> > where nALFS dies, but not if a child dies.
>
> Yes, using gdb to trace a program that forks takes some effort, it is
> not straightforward. gdb can be told to trace into children ("set
> follow-fork-mode child"), but then if it does follow into a child it
> will frequently stop tracing when the child exits.
>
> The best technique I have come up with is to insert a sleep(60); at the
> beginning code of the child I want to debug, then wait for nALFS to get
> to that point and manually invoke gdb to attach to the sleeping child
> process. This is not easy either, but it does work.

This sounds familiar. Developing GUI / multithreaded programs, I add a getline 
or something similar to the beginnig - just to attach the debugger.
But doing that for configure - no, I have no idea, how to figure that out.
When configure failed telling that it could not determine the size of long 
double, it said, more details could be found in config.log - but the end of 
the logfile looked like the beginning of a c-header-file. - So I had to guess 
again.
That's not really satisfying.

I stil presume, that I am the bug - but that's not easier to debug :(

Anyway - thanks for your hints.

Kind regards


Reinhard




More information about the alfs-discuss mailing list