Re: [Zope] Zope-killer: zSQL method that crashes Zope
If it matters to anyone, Andy isn't the only one with Zope killing ZSQL Methods I just spent way too much time trying to figure out why our stage enviroment was crashing. Turns out to be a ZSQL method which fails. Of course it wasn't wrapped in a try/except . Oh well, so much for standards...! DR Andy Gates wrote: Here's a thing. Those problems I was having with some sort of Zope-killing page have been tracked down to a particular zSQL method. The zSQL method is duff, in that is should fail - the perils of late-night coding on decaf, I fear - but it should fail neatly with a Zope error, not crash out the Python interpreter and force some poor schmuck to restart the limping Zope process. I don't know whether this is a peculiarity to my setup (NT4SP6, IIS4, Zope2, Access97 database) or a general problem; perhaps some of the developer-hatted Zopistas could take a look? The poison SQL is this: select (dtml-var field, field2) from tablename where (field1= dtml-sqlvar parameter1 type=string and dtml-var field '') Where two parameters are passed to the method, namely "field" and "parameter1". parameter1 is fed in as a dtml-sqlvar and searches against the table's field1 column. field is passed in as a dtml-var and is used to define which of the table's other fields is retruned. field2 is always returned. What makes the SQL duff is the parentheses around the fields in the first part of the SELECT. They're not required and if you remove them to produce the zSQL below, it works fine. select dtml-var field, field2 from tablename where (field1= dtml-sqlvar parameter1 type=string and dtml-var field ' So why does it kill Python? I'm only vaguely concerned now that I've found the fault, because of course the SQL is canned and no developer could get their hands on it - but this experience redoubles my acceptance of the notion of canned SQL! AndyG "I don't suffer from stress. I'm just a carrier." ___ Zope maillist - [EMAIL PROTECTED] http://lists.zope.org/mailman/listinfo/zope ** No cross posts or HTML encoding! ** (Related lists - http://lists.zope.org/mailman/listinfo/zope-announce http://lists.zope.org/mailman/listinfo/zope-dev ) ___ Zope maillist - [EMAIL PROTECTED] http://lists.zope.org/mailman/listinfo/zope ** No cross posts or HTML encoding! ** (Related lists - http://lists.zope.org/mailman/listinfo/zope-announce http://lists.zope.org/mailman/listinfo/zope-dev )
[Zope] Zope-killer: zSQL method that crashes Zope
Here's a thing. Those problems I was having with some sort of Zope-killing page have been tracked down to a particular zSQL method. The zSQL method is duff, in that is should fail - the perils of late-night coding on decaf, I fear - but it should fail neatly with a Zope error, not crash out the Python interpreter and force some poor schmuck to restart the limping Zope process. I don't know whether this is a peculiarity to my setup (NT4SP6, IIS4, Zope2, Access97 database) or a general problem; perhaps some of the developer-hatted Zopistas could take a look? The poison SQL is this: select (dtml-var field, field2) from tablename where (field1= dtml-sqlvar parameter1 type=string and dtml-var field '') Where two parameters are passed to the method, namely "field" and "parameter1". parameter1 is fed in as a dtml-sqlvar and searches against the table's field1 column. field is passed in as a dtml-var and is used to define which of the table's other fields is retruned. field2 is always returned. What makes the SQL duff is the parentheses around the fields in the first part of the SELECT. They're not required and if you remove them to produce the zSQL below, it works fine. select dtml-var field, field2 from tablename where (field1= dtml-sqlvar parameter1 type=string and dtml-var field ' So why does it kill Python? I'm only vaguely concerned now that I've found the fault, because of course the SQL is canned and no developer could get their hands on it - but this experience redoubles my acceptance of the notion of canned SQL! AndyG "I don't suffer from stress. I'm just a carrier." ___ Zope maillist - [EMAIL PROTECTED] http://lists.zope.org/mailman/listinfo/zope ** No cross posts or HTML encoding! ** (Related lists - http://lists.zope.org/mailman/listinfo/zope-announce http://lists.zope.org/mailman/listinfo/zope-dev )
Re: [Zope] Zope-killer: zSQL method that crashes Zope
Andy Gates wrote: The poison SQL is this: select (dtml-var field, field2) from tablename where (field1= dtml-sqlvar parameter1 type=string and dtml-var field '') Now I guess our code wasn't like this, but I do remember that we've had similar problems with lethal SQL methods that cause Zope to die silently and leave nothing in the logs. However, our setup was quite different: Zope 2.1.4-2.1.6, Redhat Linux, MySQL The problem is that ZSQL Methods don't seem to be very robust, to the point of actualyl killing Zope rather than returning an exception. I'd guess this is something that needs to be looked at but it's probabyl one of those bugs that's hideous to track down (ZCAtalog index bug anyone? ;-) ah well, Chris ___ Zope maillist - [EMAIL PROTECTED] http://lists.zope.org/mailman/listinfo/zope ** No cross posts or HTML encoding! ** (Related lists - http://lists.zope.org/mailman/listinfo/zope-announce http://lists.zope.org/mailman/listinfo/zope-dev )
Re: [Zope] Zope-killer: zSQL method that crashes Zope
Chris Withers wrote: Andy Gates wrote: The poison SQL is this: select (dtml-var field, field2) from tablename where (field1= dtml-sqlvar parameter1 type=string and dtml-var field '') Now I guess our code wasn't like this, but I do remember that we've had similar problems with lethal SQL methods that cause Zope to die silently and leave nothing in the logs. However, our setup was quite different: Zope 2.1.4-2.1.6, Redhat Linux, MySQL The problem is that ZSQL Methods don't seem to be very robust, to the point of actualyl killing Zope rather than returning an exception. I'd guess this is something that needs to be looked at but it's probabyl one of those bugs that's hideous to track down (ZCAtalog index bug anyone? ;-) I have a strong suspicion that nearly all bugs like this are the result of a stack overflow. It may be worth the effort to either: (1) Create a modified Python that would catch stack overflows before they happen (2) Get Zope running on microthreads/stackless Python (has anyone tried?) and gain a little more control over the stacks. Shane ___ Zope maillist - [EMAIL PROTECTED] http://lists.zope.org/mailman/listinfo/zope ** No cross posts or HTML encoding! ** (Related lists - http://lists.zope.org/mailman/listinfo/zope-announce http://lists.zope.org/mailman/listinfo/zope-dev )
Re: [Zope] Zope-killer: zSQL method that crashes Zope
Shane Hathaway wrote: I have a strong suspicion that nearly all bugs like this are the result of a stack overflow. It may be worth the effort to either: (1) Create a modified Python that would catch stack overflows before they happen Couldn't the code that catches infinite recursion be bent to do something similar? (2) Get Zope running on microthreads/stackless Python (has anyone tried?) and gain a little more control over the stacks. Way over my head ;-) cheers, Chris ___ Zope maillist - [EMAIL PROTECTED] http://lists.zope.org/mailman/listinfo/zope ** No cross posts or HTML encoding! ** (Related lists - http://lists.zope.org/mailman/listinfo/zope-announce http://lists.zope.org/mailman/listinfo/zope-dev )
