RE: Weird problems with delivered app

The only thing in your message that caught my attention was "it accesses a
database at build time". We've been told by Xanalys that dumping an image
once an odbc connection has been initialized is unsupported. Doing so on
Windows 2000 yields processor faults if the dumped image attempts to open a
db connection.

I've no idea if this is similar to your situation. You say the db is
disconnected and "pointers removed" (don't know what this means);
disconnecting isn't enough apparently, at least for a dumped image.

I hope this was of some use.

Cheers,

--
------------------------------------------
David E. Young
dyoung@bloodhoundinc.com
http://www.bloodhoundinc.com
http://lisa.sourceforge.net

"But all the world understands my language."
  -- Franz Joseph Haydn (1732-1809)



-----Original Message-----
From: Espen Vestre [mailto:ev@netfonds.no]
Sent: Thursday, June 03, 2004 9:29 AM
To: lisp-hug@xanalys.com
Cc: larsi@netfonds.no
Subject: Weird problems with delivered app


Hi,

we're experiencing really, really weird problems with a delivered
application on _two_ machines. These machines runs debian testing
and unstable but problems persists before and after updates, and 
on other machines,neither stable nor unstable do not show the same 
weirdness. Also, different motif versions in combination with 
LW 4.2.7 or 4.3.7 doesn't seem to help. Sometimes the CWD helps 
the application running a little longer (!) before it segfaults.

A characteristic feature of this application is that it accesses
a database at build time. We disconnect the database and remove
pointers (we think) before actually delivering.

This email message is for the sole use of the intended recipients(s) and may contain confidential and privileged information of Bloodhound Software, Inc.. Any unauthorized review, use, disclosure is prohibited. If you are not the intended recipient, please contact the sender by reply email and destroy all copies of the original message.

Re: Weird problems with delivered app

"Young, David" <deyoung@bloodhoundinc.com> writes:

> The only thing in your message that caught my attention was "it accesses a
> database at build time". We've been told by Xanalys that dumping an image
> once an odbc connection has been initialized is unsupported. Doing so on
> Windows 2000 yields processor faults if the dumped image attempts to open a
> db connection.
>
> I've no idea if this is similar to your situation. You say the db is
> disconnected and "pointers removed" (don't know what this means);
> disconnecting isn't enough apparently, at least for a dumped image.
>
> I hope this was of some use.

Thanks. Yes, I'm aware that database access at build time _could_ be
a reason here. However, it drives me nuts that it only very rarely
(machine-wise) causes problems. We're not using odbc, btw (but
native oracle-libs).
-- 
  (espen)

Re: Weird problems with delivered app

Espen Vestre <ev@netfonds.no> writes:

> Thanks. Yes, I'm aware that database access at build time _could_ be
> a reason here. However, it drives me nuts that it only very rarely
> (machine-wise) causes problems. We're not using odbc, btw (but
> native oracle-libs).

We have now solved this by a somewhat tricky workaround that does our
necessary compile time db access, but avoids compiling when
delivering, and the problems seem to have gone. So it seems like db
connections durning delivery was indeed the problem. We're still a
little puzzled, though, since the symptoms only occured on a few
machines, and since these machines have worked perfectly with the same
program for several months. So it would be great if Xanalys could
confirm that db connections during build is indeed a no-no in our case
too.
-- 
  (espen)

Re: Weird problems with delivered app

Unable to parse email body. Email id is 2417

Re: Weird problems with delivered app

davef@xanalys.com writes:

> FWIW without knowing the details of your build, I suggest a two-phase
> build where phase one does the database connection and compiling and
> puts the data into a file, and phase two starts Lisp again, loads the
> data from file and calls DELIVER.

yes, in fact we considered making this fully automatic, by setting a
flag that says "needs rebuild", and before calling deliver, we would
check this flag and decide whether to start lisp again or do the
deliver... btw: Is there a built-in way to programmatically 
discover that lisp is initializing? 
-- 
  (espen)

Re: Weird problems with delivered app

Unable to parse email body. Email id is 2425

Re: Weird problems with delivered app

davef@xanalys.com writes:

> There's nothing exported. I'm not sure how you'd get to use such an
> API while the system is busy starting up! What's the issue?

I'd like to know that "I'm doing a build", but since there's
no real difference between normal initializing and initializing
for a build, it's a bad idea anyway. Better to put it in the
build scripts...

-- 
  (espen)