First thoughts with 8 Beta 1

[MikeGale]

I just installed the Beta 1, here's some initial observations:

1) For a Beta install, I planned ahead and prepared the way to replace the 7 DLL in the system directory after installation. It would save time to have a switch to not install it for the Beta setup.

2) It struck me how useful it would be to have a word wrap switch on the results pane tabs. Then I could see the whole of errors, warnings etc. without needing to bring each into the message window. If this is an easy toggle I'm still free to get the present view.

3) I experimented with nsgmls as follows: options...nsglms-tab...help. In other words from the nsgmls tab I brought up help. In this state the app froze. I couldn't close, minimise or resize the help window, I couldn't get the CSE window to show, despite numerous attempts. I eventually got out by switching to another application, then back to CSE. When I came back CSE was working. I attempted to reproduce this but it didn't recur.

4) nsgmls suggested an -e switch to show more than the default 20 messages. I saw not obvious way to set that switch.

5) In my one test I found the nsgmls messages largely duplicated those from CSE, but the text was less helpful. (There was also a message that was incorrect!) I've now switched it off and don't expect to use it again unless I come across a good reason.

[MikeGale]

I've just checked what nsgmls is about.

It's the checker behind both the W3C and WDG on-line tools. Produced I think by James Clark.

That immediately tells me that it has at least one use with CSE.

I can use it to see what those other checkers will detect. (Without going on-line.)

It would be more useful if I could automatically filter out the things already detected by CSE.

[MikeGale]

I just ran another check and got an exception.

I loaded a page from a local (IIS 6) web server.

The page loaded fine, but when I came to validate it crashed. (This is a Windows 2003 machine)

Window was titled:
\rtf\b Line:\b01 cse80.exe Application Error
and message was
Instruction at 0x00200074 referenced memory at 0x0.....0. The memory could not be written.

On the first repeat cse8 simply closed down when I attempted to validate. (No message.) Also the open pages had been lost when I came into CSE.

Subsequent attempts and it validated. No errors found, no warnings, two messages and all links passed.

Later attempts (CSE closed then restarted) to validate that document also failed with a silent closure of CSE. The page is ANSI encoded XHTML 1 Transitional.

[Albert Wiersch]

Hi Mike,

Thanks for the suggestions. If you have found a crash that you can reproduce, then please tell me exactly how and send any "crash documents" to me at support at htmlvalidator dot com.

Unfortunately the components I'm currently using for the messages don't support word-wrapping. In the future, this may be an option if I switch to a different component. I'd like to eventually support this and it has been on my "to do" list.

You're right that nsgmls is like what some of the other validators use. Most people probably won't use it, but there are some people who wish to go strictly by the standards, even when it may not be best to when real-world issues are considered. Filtering the nsgmls messages so that it shows only what CSE HTML Validator doesn't find would probably be complex and bug-prone. I'm curious though, what was the incorrect message that nsgmls generated?

As for the -E option for nsgmls, there is a value name "GeneralArguments" in the key "HKEY_CURRENT_USER\Software\AI Internet Solutions\CSE HTML Validator v4\Validator DLL\Nsgmls" that you can change this. I will try to get it into the Nsgmls tab in the Validator Engine Options to make it more obvious.

[MikeGale]

Hi Albert,

Unfortunately the problem is intermittent, so not reliably reproducible. (Timing? First Page checked?) I was disappointed when it happened again. The repeats were silent shutdown of CSE 8 so I don't know of any documentation that could throw light on the issue. (I got the more useful message only once, the first time.)

The same page validated without issue in CSE 7.

The nsgmls spurious error detected a missing /head tag (?) on a page that had a normal looking outer structure, to visual inspection. (Also CSE didn't detect any such error!) Presumably something else threw it out of kilter.

[Albert Wiersch]

Hi Albert,

Unfortunately the problem is intermittent, so not reliably reproducible. (Timing? First Page checked?) I was disappointed when it happened again. The repeats were silent shutdown of CSE 8 so I don't know of any documentation that could throw light on the issue. (I got the more useful message only once, the first time.)

The same page validated without issue in CSE 7.

The nsgmls spurious error detected a missing /head tag (?) on a page that had a normal looking outer structure, to visual inspection. (Also CSE didn't detect any such error!) Presumably something else threw it out of kilter.

Thanks. The silent shutdowns are disturbing. Please let me know if any additional information comes your way. How often would you say this is happening (like x times out of y validations)?

If you come across a page that you think nsgmls is generating wrong results, then please send it to me. While I don't fix nsgmls bugs, I'm curious as to why it would do this. There is probably a reason it is missing something. I suspect it's not a bug.

[MikeGale]

In my brief test the corresponding messages had regular and corresponding text descriptions, this suggested a simple technique:

Code: Select all

If both nsgml and cse messages on line

Extract signature from each nsgml message

Match signature with all cse messages

If match dump nsgml message

[Albert Wiersch]

In my brief test the corresponding messages had regular and corresponding text descriptions, this suggested a simple technique:

Code: Select all

If both nsgml and cse messages on line

Extract signature from each nsgml message

Match signature with all cse messages

If match dump nsgml message

Thanks Mike. That sounds pretty good. I may have to experiment with that at some point, especially if a lot of people find the nsgmls integration helpful and use it often.

[MikeGale]

Hi,

I've looked at the failing page again. (page sent in separate email) Details:

1) CSE 8 installed but CSE 7 DLL in system directory and CSE 7 installed in parallel.

2) The page now fails every time whether it's the first or a subsequent page. Formerly see previous posts.

3) The page was created as follows: Framework designed and tested, page published, content edited by client from different web site using CMS tool that emits valid xml (not exhaustively tested for xhtml compliance), edited content inserted into original document, content published back to original site (FTP). Page validates with CSE 7.

4) Test configuration. CSE 8 running on same machine as IIS 6 server that serves up page. Web site is mainly *.htm with a few *.asp pages, none of which were invoked in the test. CSE 8 configuration imported from CSE 7. Windows 2003 Ent, service packed up to date (yesterday was last change, a bunch of maybe 8 items).

(Interesting point. My first publication of the algorithm post vanished into the great bit bucket in the sky while I was previewing. I think I posted it pretty much the same time as your reply post and the forum software failed the transaction test.)

[Albert Wiersch]

Thanks Mike,

I have tried to reproduce the problem with the file you sent but couldn't. Do you have HTML Tidy or nsgmls messges turned on? Do you have accessibility checking on? If so, what settings?

Also, could you send me the config file you're using with v8.0 so I can try it with that? I assume you're using a modified v7.0 config file.

Thank you.

[MikeGale]

Hi,

To get a clean slate I ran a reboot (something I normally avoid) and tested again. Findings as follows:

1) After reboot I think I got three messages from OS as follows, "I shut down do you want to report", "I closed CSE" and "I closed CSE do you want to report".

2) After that validation worked, maybe 7 times. I ran it immediately after a fresh start of CSE, and after validating another page first. All of these are from web and not from filing system.

So the problem has gone away. To me it seems very odd indeed. It had affinity for one page and one page only, among those I tested. It was reproducible. I'd like to think it was rare but it did hit me after very little testing, so I wouldn't bet on it.

On the other questions, in case this rears it's head again, no Tidy Messages, no nsgmls messages and no accessibility checks when failing. Yes the configuration was based on a v7 configuration, but one additional message had been disabled from the "click on message pop up menu". (The issue may have occurred immediately after nsgmls was disabled. The first occurrence was unique in that it gave an exception message.)

I'm really glad that's out of the way, for now.

[Albert Wiersch]

Hi Mike,

Yes, that is very bizarre, though I don't understand what you mean by:

1) After reboot I think I got three messages from OS as follows, "I shut down do you want to report", "I closed CSE" and "I closed CSE do you want to report".

I am using a new system that should give you the option to send reports about crashing, so maybe this is asking you those questions? I don't know why the OS would display those? Can you clarify this and tell me more about those messages?

Also, you should have been given the chance to report the crash instead of CSE HTML Validator just disappearing or exiting silently. So something went wrong here too, but I don't know what. I am wondering if any security software you may be running could be causing problems. It seems that as security software gets more complicated and tries to find more stuff it seems to cause more issues.

[MikeGale]

Hi Albert,

Those messages are a Windows 2003 thing. Just part of the normal run of operations, after reboot. I seldom send the report but I really like a reminder of anything that might have gone wrong. (I've never taken the time to get a deeper insight into this subsystem.)

I'm running a fair amount of "fight the slime" processes, firewall, anti-virus and SysInternals Process Explorer come to mind. As you say they can interfere. I never saw your "send a report" interface.

I've just had a thought, if it happens again I'll see whether I can retrieve details via that prompt.

[Albert Wiersch]

OK, thanks.

I wouldn't be surprised if one of those "anti-slime" programs was helping to cause the problem or prevented the CSE HTML Validator crash reporting mechanism from appearing.

Please keep me posted!

[MikeGale]

Hi,

If it happens again I'll try to dig out more information.

Currently I have a dump of knowledge base articles that Microsoft automatically selected from the logged event. It's not very specific as it lists what I estimate is more than 90 articles. I put these into a word document. (There was too much for me so I didn't go any deeper!)

If you'd like it I'll send it through. let me know. (Each hyperlink has a one paragraph description.)