2.0.1 (final) Bug Report Thread

[mikelove]

sthubbar - to be honest, the Manage Flashcards screen is such a user interface morass that rather than making more incremental improvements like this I'm inclined to just overhaul the whole thing in 2.1 - I mentioned the idea of an "Organize Flashcards" screen elsewhere, we can make that the primary interface for things like reassigning categories and turn Manage into something mainly for complicated searches / batch operations. With the flashcard-less first iPhone version, it looks like 2.0.2 is going to be light on flashcard improvements in general, though there should be a couple of nice little enhancements on the dictionary side of things.

Dictionary sort order is stored in internal memory exactly like Pleco's other preferences, but it's possible that it might get screwed up by that dictionary change / reinstall; the preferences should be retained even when a particular dictionary temporarily fails to load, but there are certain circumstances under which we need to dump them.

The need to store flashcard / user dictionary databases in hidden internal memory is unavoidable, unfortunately - putting them in PDBs just isn't reliable enough, Palm's NVFS filesystem is way too buggy (particularly on an older model like the Treo 650). Having your flashcard database randomly corrupted in a way that you might not immediately notice (and which might therefore infect your backup copy as well) would be even worse than losing it due to forgetting to back it up.

But this issue is noted in several different places in the instruction manual, including the 2.0 upgrade guide ("Important Note for Palm OS Users") and the Palm OS installation guide ("Important Note" right at the top), and I also made specific reference to it in the 2.0.1 announcement e-mail, so short of having a loud voice say "BACK UP YOUR FLASHCARDS MANUALLY" every time people launch Pleco I don't know what we can do to make it clearer

ipsi - that's odd... sounds like Windows is resizing the display without telling us it's done so. Should be easy enough to fix, though, just look for the WM_SIZE window-size-changed message and either restore or register that we're no longer in fullscreen mode if the window is being resized to something other than fullscreen.

 

[gato]

But this issue is noted in several different places in the instruction manual, including the 2.0 upgrade guide ("Important Note for Palm OS Users") and the Palm OS installation guide ("Important Note" right at the top), and I also made specific reference to it in the 2.0.1 announcement e-mail, so short of having a loud voice say "BACK UP YOUR FLASHCARDS MANUALLY" every time people launch Pleco I don't know what we can do to make it clearer

Maybe have a message that pops up once a week on startup that reminds people to backup their flashcards? The pop-up message can have a checkbox that allows people to turn it off.

 

[mikelove]

Maybe, but I'm not wild about annoying messages, even ones that can be turned off - if we were going to do something timed with backups I'd rather it be an automatic backup-to-SD-card or something like that (perhaps zipping the resulting file to save space), though at this point that's only really feasible on Windows Mobile.

 

[sthubbar]

I'm pretty sure the pinyin v processing has some issues.

I was randomly noticing the problem and at first thought maybe I had typed something wrong. Now, I have seen it so many times and on cards that I know I recently input that I'm positive that there is something wrong going on.

I have seen this on cards that have been imported from 1.0 as well as cards that I enter manually within 2.0.1.

I'm pretty sure these are all custom flashcards.

It is also possible that these are all long cards/sentences. I'm not sure if that is relevant because almost all of my cards are long.

The pattern seems to be if there is a third tone v such as in 女, I enter nv3 and then later on it gets changed to n53.

This is even weirder in that I was just reviewing a card and the pinyin looked like nu(with umlaut)3. So the u with umlaut had no tone over it just two dots. Hope that is clear, I saw three characters, the n, the u with umlaut, and the number 3. Now what is weird is I went in to edit the card and in the card I saw n53. There was no v, and no u. I don't know where that u with umlaut came from or went to, but it disappeared.

 

[mikelove]

Have you turned Input Field Compatibility on or are you still running with that turned off? Having that option turned on could definitely cause this problem, since the character code for an umlaut conflicts with the codes used by GB / Big5 to represent Chinese characters.

If IF Compatibility is off, I'm not sure what else the problem might be - is there anything specific about the location of this syllable in the word that seems to bring on the problem? Beginning or end or middle, other words around it, spaces, etc?

 

[ben_gb]

Hi Mike,

Two things (bugs, maybe) I've noticed in the flashcards...

1. If I am doing a flashcard test, but then get diverted and my phone locks, when I unlock it the top line doesn't get updated (it shows the start button and time, instead of the card stats).

2. I use the 1-6 numbers option for how well I know the cards. If I also select the 'review incorrect words at the end' option, when I come to the end and review the cards I got wrong, it just shows the tick or cross button, and not the 1-6 numbers. Is this intentional?

Regards,
Ben

 

[sthubbar]

I have input compatibilty mode turned on. I don't know any other pattern than it seems to occur with third tone.

 

[mikelove]

ben_gb - #1 is probably related to that keyboard redraw issue from before, the software isn't being notified that the screen has been resized for some reason.

The lack of 1-6 in the end test is indeed intentional - there's no point in collecting them since the cards' scores aren't being recorded, the only reason we're still collecting correct / incorrect answers at all is for the "loop" option so you can keep seeing the incorrect ones until you get them all correct.

sthubbar - which "External Text Encoding" are you using, UTF-8, GB, or Big5? I have an idea for something that might be causing this if you're using Big5, but I'm not as sure about the other two.

 

[sthubbar]

I am using GB, with input compatibility enabled, and I have found a repeatable example of the problem.

Put the following text as the pinyin of a custom flashcard, the hanzi doesn't matter. Then after saving the card you should be able to immediately go back and look at the card and it should be weird. The display will show nu(with umlaut)3, but if you go in and edit the card you should see n53

zhe4shi4yi2ge4shan4liang2ke3ai4denv3wu1

BTW, I just tried disabling input compatibility mode and there was a change, though the problem is almost exactly the same. The only difference with input compatibility mode disabled is that i I go in and edit this new card I see n5u(with umlaut)3. With input compatibility enabled, I just don't see that u(with umlaut).

 

[epic709]

Hi Mike,

I just upgraded to 2.0.1 yesterday on my aging Palm T5. I'm pretty impressed with all the changes (I started using Pleco Ox-Dict 5 years ago on my CLiE and absolutely loved it as a Chinese beginner learner). However, the Reader module does not work for me when I try to read from "PalmDoc File (Internal)" and it consistently soft-resets with the error "MemoryMgr.c, Line:3759, Non-word-aligned handle" (see attached image). Searching on the forums, I found 2 other threads reporting a similar but not exact error (different versions of Pleco and not on Palm T5). Reading those threads, I get the impression that this bug was so rare that a fix was deemed unnecessary.

I was really looking forward to extending the usefulness of my T5 by purchasing Pleco 2.0.1 and all it's new add-ons so appreciate any support you can provide.

 

[mikelove]

sthubbar - that explains it; looks like an obscure bug in our Pinyin conversion system, "den" is actually a valid Pinyin syllable and the code that disambiguates cases like this (chan gan versus chang an, etc) by requiring an apostrophe / space before the start of a new mid-word syllable that begins with a vowel doesn't seem to be recognizing the umlaut / 'v' as a vowel.

Thanks for following up on this - should be easy to fix in 2.0.2, and in the meantime if you insert a '5' or a '0' after the 'de' (to indicate a neutral tone) that should get the Pinyin to be interpreted correctly.

epic709 - it looks from that screen like there aren't any actual PalmDocs installed on your Palm; the software still shouldn't be crashing in that case, but to actually use the document reader you'll need to load up some documents to read. You can try it out using the "Send to Reader" menu command - that will bring up the current dictionary entry in the reader - or by copying some Chinese text to the clipboard and choosing "Clipboard" from the Source menu.

An excellent source for Pleco reader documents is stephanhodges/ipsi's website at http://china.panlogicsoftware.com/ebooks.html - you can download a PalmDoc file from there and install it right to your Palm.

 

[hejincong]

Sadly, I'm still having trouble getting Pleco to work with my HP iPAQ 614c's 10-digit hardware keyboard. When I tap the number keys, instead of being interpreted by xT9, I only get the numbers.

I think there might be a simple registry edit that could fix this, however. I had the exact same problem with Opera Mobile 9.5 until I found this post:

http://forum.xda-developers.com/showthread.php?p=3357126#post3357126

It adds a "CustomClassName" in order to have xT9 recognize it. Is there a value for Pleco that could have the same result?

 

[sthubbar]

Mike, I am seeing some unexplained behavior with full text searches.

Hmm, I was just trying to reproduce the problems I noticed before, and now the situation is different.

"#roll " [space] at end produces an entry in the TL dictionary or ABC where it says "roll of"
if I press o so searching for "#roll o" I receive the "No matches" screen.

What is messed up is just earlier today, I was searching for

"#roll eye" and I was receiving a 21 dictionary match that had 'roll eyes' in it, when I would type
"#roll eyes" just by adding an s it would say 'No matches"

Now when I search for '#roll eye' or even "#roll e" I am getting "No matches"

Maybe I was doing a different search.

 

[sthubbar]

Hmm, I thought I figured part of this out that maybe the full text search only worked with words. I thought this because when I searched "#trave" I receive "No Matches" yet when I search "#travel" I receive a match. Hmm, but then it doesn't make sense because "#tr" returns the entry 21C Am.trak. "tr" does not appear to be a word so why does that return and not the trave in travel?

How about the "#girlfriend off in a" is no good "#your girlfriend off in " works so it doesn't look like the search is too long. This is more than most people would want though maybe it is showing a weakness that might be affecting simpler searches.

 

[mikelove]

hejincong - it could help, yeah - try adding "PlecoFieldClass" to that same "CustomClassNames" key and see if that helps. Also add "PlecoSelectorClass" if you want to be able to use it for flashcard free-answer inputs.

sthubbar - full-text searches are only designed to be word-based; they're not designed to work with partially-completed words, though in some cases thanks to a combination of factors they may still turn up results for those. "roll o" is failing because "o" appears in one or two places in the ABC dictionary as a whole word and Pleco isn't finding an entry where "roll" and "o" both appear (or at least not one where they appear next to each other). "tr" is coming up with "Amtrak" in that strange way because the code which searches for a range of characters to highlight is rather dumb and doesn't worry about whole words - a few lines down in that entry you'll see a "tr(avel)" which is where it's getting the whole word for "tr". "girlfriend off in a" I'm not sure about, it may be running out of memory and aborting the search (too many lists of entries containing a given word to synchronize) - perils of Palm development. At some point we'll likely roll out a separate interface for full-text searches with a dedicated "search" button, but for now you should just make sure to type out each word in full.

 

[hejincong]

hejincong - it could help, yeah - try adding "PlecoFieldClass" to that same "CustomClassNames" key and see if that helps. Also add "PlecoSelectorClass" if you want to be able to use it for flashcard free-answer inputs.

It works! Not perfectly, but it works. Usually I can use the directional buttons to scroll between T9 options, but Pleco captures the keypresses. I'm not complaining though, it's already improved more than I'd hoped. Thanks a million!

 

[mikelove]

Great! I'll have to mention this to other T9 users then.

 

[jiacheng]

I have noticed that the statistics seems to calculate the incorrect number of cards in a category. I'm thinking that the algorithm must not look for words that are contained in multiple sub-categories of the parent category being used to calculate the statistics. For example, if you were to search my categories uncer PCR 1, you the flashcard manager shows 566 cards. Where as the statistics page calculates 605 cards. What's more is, it seems to assume that all 39 these phantom cards are all 0-100 score range.

 

[anchan42]

Flash card advanced setting in cards page, every time I change anything in dropdown list the "Unlearned card ratio" line seems to flash.

The last letter on some of the control captions in advanced setting have got the last letter partially covered. This make an n looks like an r.

 

[mikelove]

jiacheng - that's odd... are you sure you're just looking for a category and nothing else under Manage Cards? That multiple sub-category theory makes sense, though I'm not sure why the cards would then come up with 0-100.

anchan42 - yeah, those got a bit messy when we redesigned them in 2.0.1 - should be easy to reduce the flickering in 2.0.2 at least.