Conference smurf::buildhelp

Date Of Receipt: 	 4-SEP-1996 16:00:06.03
From: 	WASTED::"[email protected]"
To: 	19.779::melvin (USEG #381-1593 @ZKO3-2/T21)
CC: 	wasted::odehelp, [email protected], help_ode@DEC:.zso.slugbt
Subj: 	Re: My bsubmit problem ( as reported recently)
Joe Melvin wrote: 	

>I appear to have worked around the problem.  A mail message from
>Josh implied that the problem might be related to the LONG audit
>trail I was using for the submit.  While bci appears to have taken
>it, bsubmit did not.  I bco/bci the two files; during the bci I
>used a much shorter audit (LOG) entry.  I then was able to do the
>bsubmit -resb etc successfully ...[snip]

>So, the problem does appear to be the length of the audit (LOG)
>trail.  In future, I will keep this entry as short as possible
>to avoid this problem.


Joe,

By way of explanation, I think you've run into a limitation imposed
by the RCS interface. What bsubmit does is to take the log comments
for *all* the revisions on your private branch and bundle them into a
single comment. Unfortunately, the only way for bsubmit to pass that
comment to RCS is through a shell command line. If the shell command
that bsubmit creates is too long, the shell truncates the line and the
command eventually returns an error to bsubmit which has no idea what
the problem was. Hopefully, we'll be able to modify the RCS interface
for a future release of ODE.

Meanwhile, you've already found the workaround answer: Keep the log
comments fairly short.

John D'Addamio

Date Of Receipt: 	 4-SEP-1996 16:01:24.91
From: 	FLUME::jmf "Joshua M. Friedman Digital UNIX"
To: 	19.779::melvin
CC: 	odehelp@DEC:.zko.flume
Subj: 	Re:  My bsubmit problem ( as reported recently)

Joe, you don't need to keep your submit logs "as short as possible",
but if you keep them to 20 or 30 lines then you'll be ok.

The DECwest ode team is investigating this and will let us know what
the actual limit is, and if it can be expanded.

-josh

Date Of Receipt: 	 4-SEP-1996 16:27:06.77
From: 	FLUME::"[email protected]" "04-Sep-1996 1325"
To: 	Joshua M. Friedman Digital UNIX <[email protected]>
CC: 	19.779::melvin, [email protected], [email protected]
Subj: 	Re: My bsubmit problem ( as reported recently)

Hi Josh,

>The DECwest ode team is investigating this and will let us know what
>the actual limit is, and if it can be expanded.

If the problem is what I think it is (i.e. the shell command interface to RCS), 
we can't detect the limit or easily expand it. The only way I can see to
change the limit is to modify RCS to take a file name for the log comments 
rather than passing them via the -m option which is what we do now.

I've put the whole issue of history logs into our design process for
the next release. Hopefully, this will resolve *all* history log problems
including this one.

John

Date Of Receipt: 	 4-SEP-1996 17:16:10.90
From: 	FLUME::jmf "Joshua M. Friedman Digital UNIX"
To: 	[email protected]
CC: 	odehelp@DEC:.zko.flume, tea@DEC:.zko.flume
Subj: 	Re: My bsubmit problem ( as reported recently)

Ok John, thanks.  This seems to be an issue around comments that are
something like 40 or 50 lines; somewhere around there is a buffer limit
of some sort.  

Last week someone hit it after many bco/bci's, and all the bci logs were
wrapped up into one (too long) submit log.

-josh

Date Of Receipt: 	 5-SEP-1996 12:47:32.37
From: 	FLUME::"[email protected]" "05-Sep-1996 0945"
To: 	Joshua M. Friedman Digital UNIX <[email protected]>
CC: 	[email protected], [email protected], [email protected],
	[email protected]
Subj: 	Re: My bsubmit problem ( as reported recently)

Josh,

>Ok John, thanks.  This seems to be an issue around comments that are
>something like 40 or 50 lines; somewhere around there is a buffer limit
>of some sort.  

I didn't mean to imply that I wasn't going to do anything further. I will
verify that bsubmit does pass the comments via the RCS -m flag and that it
is the shell's limitations that gouge us. Beyond that, the best I can do is
experiemtn and give you an approximate max number of *characters* that can
be in the comments before the shell truncates. I say approximate because
it may well vary from machine to machine or release to release of the OS.

John