FWIW, using the exec trick has made the mysterious premature
termination go away. Much cleaner.
I'm going to let the current build finish and then check in this
change for future generations!
Thanks, Don!
On 2008-07-15, at 10:30EDT, P T Withington wrote:
Ah. That seems like a better approach. Did not know about the exec
trick. Thanks!
On 2008-07-15, at 10:12EDT, Donald Anderson wrote:
Oh yes, you're right. I guess the only downside with yours is if
there was another 'here' document embedded.
Or as you say, if there's a size limit (though I would think it
would allow pretty large ones).
Hmm, also in your version I don't think any variables ($foo) known
to the outside shell
get substituted.
Yet another approach that avoids a subshell and should be even
closer to the original:
set -xv
exec >> log_file 2>&1
....script body...
(lots of examples of using exec to manipulate file descriptors here:
http://www.linuxtopia.org/online_books/advanced_bash_scripting_guide/x13082.html)
On Jul 15, 2008, at 9:54 AM, P T Withington wrote:
I think what I wrote and what you wrote boil down to the same
thing, don't they? Create a subshell with verbose and expand on,
send standard and error output to a file, run ...script body... --
I supply the script body as a 'here doc' on standard in, you are
supplying it 'in line' to the subshell. Under the covers who
knows if there is a difference.
But, since I am at a loss, I think I will try your formulation to
see if it works any better. I wonder if I am hitting some stupid
limit on the size of a here doc. Seems very un-gnu-like, but who
knows...
On 2008-07-15, at 09:38EDT, Donald Anderson wrote:
Won't the technique you're using just put the text of the script
into the log file?
(instead of executing the script).
Several ways around this, easiest might be:
(
set -xv
...script body....
) >> log_file 2>&1
On Jul 15, 2008, at 9:21 AM, P T Withington wrote:
I tried to make nightly-go log more of its output by taking the
main body of the script and executing it in a subshell like so:
sh -xv >> ...log file... 2&>1 <<'EOF'
... script body ...
EOF
The idea is that the log file will see the actual steps of the
bash script, not just the output of the ant_call. If I comment
out the step that fetches the source (just as a test), then the
script runs just fine through all the steps and I get the
expected failure message. But if I do an actual build, right
after the dist-one build finishes, the whole script just stops
(even though there are many more steps).
I'm stumped. If no one has any suggestions, I guess I will
remove this bright idea of trying to capture all the output and
give up.
--
Don Anderson
Java/C/C++, Berkeley DB, systems consultant
voice: 617-547-7881
email: [EMAIL PROTECTED]
www: http://www.ddanderson.com
--
Don Anderson
Java/C/C++, Berkeley DB, systems consultant
voice: 617-547-7881
email: [EMAIL PROTECTED]
www: http://www.ddanderson.com
