[ic] installing IC 4.8.2: various problems
Steffen Dettmer
interchange-users@interchange.redhat.com
Fri Nov 2 21:11:00 2001
Hi,
I write jsut ONE mail for serveral issue. Dear reader, just skip
the sections you're not interested in. I would be really glad if
you heard about similar problems; because of that I tried to
supply many information. Thank you in advance for your time.
just tried 4.8.2. First, I merged my changes with the foundation
demo. According to UPGRADE, I expected no problem. But, well...
- DATABASE issues:
------------------
First, I would really like to chenge the default to not drop all
known database tables (of course I forgot to set the NoImport
thing). I say: no, don't install sample data and got empty
tables. I feel that's bad; if nothing should be inserted, nothing
should be deleted. Maybe an additional option in the makecat
would be possible (like: warning, existent, non-empty tables
found, do you really want to screw up all your data?).
- Installation (configure):
----------------------------
Well, the CPAN bundle installation failed of course, since I
don't allow the web server (which is running IC) web access. I
was sure that I installed the Bundles on the last installation
(as root system wide as I though). Well, I was wrong. I allowed
proxy and things, used -MCPAN -eshell to install the Bundle. I
saw some errors happily leaving my terminal on fast scrolling. No
chanche for interaction. Well, after it IC was still not working.
Well, I had now a installation in /tmp/interchange. Hum. My perl
installation is not the best... Of course this is not in @INC.
Hum. Nice install default anyway, cp'd it into interchage/lib.
Was this a good idea? At least IC started. In the meantime I
restored the database with the lastest backup and IC came up with
my catalog. Puh.
INET Mode with non-default port
-------------------------------
To be able to run both old and new version, I incremented the
inet port to use by one. I would like that the tool would print a
hint, that the interchange.cfg has to be changed manually (the
errors are misleading after that, and not anybody is so paraniod
to try netstat -an --inet | grep LIS at first :)).
- set proc title hack?
--------------------------
Well, I had some messages (on some SuSE, BTW) and read that it
may be cause due to setting $0. Well, proctitle things are not
official supported in linux, maybe perl overwrites it's argv with
memset, well, not nice but working and not my business, I removed
$0= in bin/interchange. Well, suddenly I had to install
File::CounterFile. Hum. Oki, cp'd from /tmp, IC launches (still),
but with no error messages.
For a server I implemented, I used such a hack, and please not,
that the kernel still knows how long the name was; ps padds it
with spaces (even if you memset zero the space!). So the output
of ps is ugly. I had half a screen spaces...
- endless loop after all:
--------------------------
On some (!) shop actions (most time it's clicking on "buy"), the
IC process gets on some endless loop (browser blocks of course).
Strace showed:
stat("/usr/local/interchange-4.8.2/lib/auto/Storable/store.al",
{st_mode=S_IFREG|0444, st_size=532, ...}) = 0
getgroups(32, [509, 514]) = 2
mremap(0x4028f000, 954368, 954368, MREMAP_MAYMOVE) = 0x4028f000
in an endless loop, some 100 times a second. Don't reacts on
TERM, but on KILL. How I can debug it? I made an strace from a
whole request, looks good, it reads the pages, queries the DB,
receives results, and after 2500 lines I get:
[pid 27204] stat("/home/bedinetz/catalogs/bedi2/session/x/m",
{st_mode=S_IFDIR|S_ISGID|0770, st_size=96, ...}) = 0
[pid 27204]
unlink("/home/bedinetz/catalogs/bedi2/session/x/m/xms4aAq4:std")
= -1 ENOENT (No such file or directory)
[pid 27204]
stat("/usr/local/interchange-4.8.2/lib/auto/Storable/store.al",
{st_mode=S_IFREG|0444, st_size=532, ...}) = 0
here starts the loop. I have no idea how to debug, seems to be
something weired in autoload or so? Any ideas?
Won't mail the trace, since it's large :)
- "new account" button in login.html fails SOMETIMES.
--------------------------
It seems that the following patch helps:
bedinetz@stella:/usr/local/interchange-4.8.2/bedi/pages # cvs
diff -u -r 1.2 -r 1.3 login.html
Index: login.html
===================================================================
RCS file:
/net/home/bedinetz/cvs.std/bedi-catalog/pages/login.html,v
retrieving revision 1.2
retrieving revision 1.3
diff -u -r1.2 -r1.3
--- login.html 2001/10/03 16:59:29 1.2
+++ login.html 2001/11/02 16:32:10 1.3
@@ -41,10 +41,11 @@
<FORM ACTION="[process secure=1]" METHOD=POST>
<INPUT TYPE=hidden NAME=mv_todo VALUE=return>
- <INPUT TYPE=hidden NAME=mv_click VALUE=Login>
+ <!-- INPUT TYPE=hidden NAME=mv_click VALUE=Login -->
<INPUT TYPE=hidden NAME=mv_failpage VALUE="login">
<INPUT TYPE=hidden NAME=mv_successpage VALUE="customerservice">
<INPUT TYPE=hidden NAME=mv_nextpage VALUE="index">
+ <input type=hidden name=mv_session_id value="[data session id]">
<table border="0" align="center">
<tr>
@@ -69,7 +70,7 @@
<tr>
<td> </td>
<td class="contentbar1">
- <input type=submit value="Log In">
+ <input type="submit" NAME="mv_click" value="Login">
<input type="submit" NAME="mv_click" value="New Account">
</td>
</tr>
@@ -92,6 +93,7 @@
[else]
<FORM ACTION="[process]" METHOD=POST>
+ <input type=hidden name=mv_session_id value="[data session id]">
<INPUT TYPE=hidden NAME=mv_todo VALUE=return>
<INPUT TYPE=hidden NAME=mv_click VALUE=Logout_choice>
<INPUT TYPE=hidden NAME=mv_nextpage VALUE="login">
I don't know if it's correct, but I think there may be problems
when there are two mv_click CGI variables?
- IC debug mode not working
---------------------------------------
I started
/usr/local/interchange-4.8.2/bin/interchange -r -DEBUG=1
and ordered something in the shop. Instead of the endless loop I
got:
bedinetz@stella:/usr/local/interchange-4.8.2/bedi/pages #
/usr/local/interchange-4.8.2/bin/interchange -r -DEBUG=1
The previous Interchange server was not running and probably
terminated with an error.
##### DEBUG MODE, running in foreground #####
Low traffic settings.
Calling UI...Interchange V4.8.2
Configuring catalog bedi2...Using PostgreSQL,
DSN=dbi:Pg:dbname=bedi-shop;host=laxx;user=interchange;password=ups_thankyou...done.
Interchange server started (30282) (INET and UNIX)
Running in foreground, OS=linux, debug=1
String found where operator expected at (eval 346) line 1, at end
of line
(Missing operator before ?)
panic: POPSTACK
Goto undefined subroutine &Vend::Error::get_locale_message at
/usr/lib/perl5/5.00503/autouse.pm line 61.
Is there something missing in the installation or so? Any hints?
I cannot understand the message. A find|xargs grep showed, that
Vend/Error.pm do define a sub get_locale_message at least.
Well, the stuff is really complex and I think it's not possible
to perl -d / ddd / gdb it in any case. Hum. Any hints?
Not possible to order checkout
------------------------------
I created my steffen user with address, and went to checkout
page. I allowed postal order to have no problems. I just wanted
to check if the mails still get send and so on. But after filling
out the form, the same page comes up, as if I had made input
mistakes, but there are no messages displayed. I get no error,
but the same page ever and ever. In the error.log there is no
message about it (but I thing this is normal, since validation
isn't loggin at all AFAIK). Any ideas?
Quantity Pricing: wrong prices
-------------------------------
Since my shop isn't working with IC4.8.2 today, I tried the RH
demo store. I set up a steffen, said it's an dealer and
configured quantity pricing for the articel os*007.Nice SKU, BTW
:). In basket, there is the link "QUANTITY PRICING". But the
prices are incorrect. First, the quantify numbers are not the
same as in the DB, and second the values are not the ones from
the DB (i.e. 25 is not the value from q25 or w25 or how the
fieldname was). Is that feature believed to work? Are there
additional things that modify the price? The value for q5 was
correct, BTW.
Thank you, dear reader for taking the time for reading my
posting. I wish a happy weekend to the list!
oki,
Steffen
--
Dieses Schreiben wurde maschinell erstellt,
es trägt daher weder Unterschrift noch Siegel.