Major problems after saving and reopening a file with numerous triplets

• Oct 26, 2011 - 00:15
S5 - Suggestion

I spent all day on a file - about 7 hours. After finally finishing and saving the file (and then saving again as a pdf), I closed the program and re-opened to make edits.

I then found that musescore has completely jumbled the project - it seems to be having major issues with all the tiplets, which are removed and scattered.... with lyrics missing and placed improperly all over the place. the file is basically useless.

Is there any way to find a prior version of a musescore file? I'm hoping I can somehow recover the file as it was before it inexplicably got ruined.

I'm attaching the pdf, which is how it should look (although I do see one error where the triplets were somehow ommitted - measure 38 at the bottom of page 3) - and also attaching the musescore file, which for me has become a giant mess. if anyone can help or advise me on something i did wrong, i'd appreciate it.

PS I know I could've used 12/8 time, but this is a job for someone else who specifically requested 4/4 since the accompanying music was arranged that way. Is it the triplets that caused the issues?

Attachment Size
I've Got a Secret.mscz 17.46 KB
I've Got a Secret.pdf 81.01 KB


Look for a file called ".I've Gor a Secret.mscz,", i.e. the same filenames but starting with a dot and ending with a comma

I too experienced a loss of triplets once. Never found why or even when that happened

Looking at your files, appart from line breaks in different places, I see the first real diff in measure 17 (no triplets and lost lyrics). Is it that you're talking about?

that's strange - maybe this is something that's only appearing on my computer? my file is totally corrupted... missing lyrics everywhere, triplets gone, and numerous added measures in irregular time (only 2 beats to a measure, etc). the problems for me begin at measure 7 and they're all over the place on page 1.

i'm totally flabbergasted. i supposed i'll have to re-create the file again using the pdf as a guide, because my client wants to make some edits and i noticed a problem or two.

but i'm very worried if i take the time to re-create it, the same thing will happen again.

i did find the comma file, but unfortunately the date says 10-20-11, which is when i first created the file but before i did any of the work. that said, how do you open those types of files?

Load it into the nightly and you will see the corruption. I was hoping to load it there and save as an XML for re-import back to 1.1 but no go.

thanks jojo - i was able to open it but it's the version from before i started any work on the file, so it's not helpful :(

do i need to download the nightly and work from there? what do you guys suggest for tackling this problem? I'm resigned to re-creating (and after going through the whole thing, there are large portions that aren't corrupted, so it's not as bad as I first thought). But i don't want this to happen again.

also i should add - this is just the first of many numbers i'm doing for this client (a vocal libretto for an entire score), and it's FULL of triplets... doing it in 12-8 is not an option because his accompaniment score is in 4-4 and he wants it to match.

so i'm perplexed. i love musescore and don't even have finale or other options, so i hope there's a solution.

I've had many problems with triplets being corrupted. For example, triplet crotchets (1/4 notes) in a 4/4 measure being changed into three straight crotchets, leaving a 5/4 measure in one part, and screwing up a whole score thereafter. Hard to deal with as Musescore doesn't accept that the 5-beat bar is an irregular measure, but won't reset it to 4/4 no matter what you do. So you end up having to delete the measure across all parts in the score, insert a blank new measure, and recreate each part by hand. Disappointing, to say the least! And yes, there are places where only triplets will do, so it's not possible to work around with alternative time sigs or notation.

I think this is a bug-fix worth prioritising if possible.

Sorry! I'm using v1.1 on Windows 7.

I know my imprecise observations aren't much use in helping developers recreate the error, but these instances of corruption (several times, but *not* all the time) happen when you've got your back turned... I haven't yet managed to catch MS in the act and see why/how it might be happening.

I'm working ( slightly nervously) on a fairly large score with triplets at different points in a few parts. Will keep you posted if I spot anything.

There have been so many possible corruption vectors posed wrt tuplets that its confusing. Copy/paste has been blamed for much of the problem, but there's more to this. I've tried creating a score with many triples, various kinds, but I couldn't make it corrupt. However, in experimenting this morning I've finally managed to find a way to create corrupt measures very easily. I don't know if this method is what could be happening to people, but its a starting point:

1. Create a new piano score, key of C, 4/4 with a 1/4 pickup measure
2. Click in the first bar and type B, F
3. Arrow back and type Ctrl-3 (on the B)
4. Type A, G, to complete the triple
5. Arrow back, then arrow forward to select the last note (G) of the triple
6. Type 5 (this selects the quarter note)
7. Type Ctrl-3

The first and second bar are now corrupted. Once this file is saved, the corruption is embedded. Note that this whole time you were in note entry mode, and if you skip step 4 (adding the real notes to the triple) the problem doesn't happen. Real notes have to exist.

The problem seems to come from allowing the changing of the note length (step 6) before a ctrl-# option as this seems to screw things up quite badly. You are telling the tuplet generator what note length you want the select note split into. Typically you only want to split it down into the next smaller note (half to quarters, quarters to eighths, etc) but by changing the default note length, the tuplet generator seems to get the wrong info on how to split.

I don't think that the pickup measure has much bearing on this problem, although it helped to bring out the error quicker. I've managed, using similar methods above, to horribly mangle a standard 4/4 score.

Great that you've isolated a repeatable case, schepers! I suspect that's worth a separate bug report, as it will make it easier to find the problem report, and also, there is no guarantee that fixing that bug also fixes the one that the OP is seeing.

Meanwhile, just some words of advice when dealing with triplets. This is based on my experience; no guarantees:

1. Save work often. If you're backup is from when you first created the score, that means you never saved it until the end? Dangerous practice in any case (not just with triplets, and not just with Musescore). Doesn't hurt to save snapshots under different names periodically.

2. Be on the lookout for any sign of corruption, and hit Undo immediately if you see one appear, rather than hoping it foes away. If you're so inclined, see if you can remember what you did to trigger the problem, try it again, and if it's reproducible, post an issue! If Undo doesn't work, then delete the measure, insert a new one, and move on. You can try copying and pasting good notes from the old to the new, but double check that this doesn't copy the corruption.

3. Because sometimes corruptions don't appear until a reload, reload periodically to see if any problems have appeared. You may be able to recover to the most recent backup or just by deleting the offending measures and moving on, without losing mich work.

4. The most problematic things seem to be copy and paste of partial measures involving triplets and creating triplets that don't start on traditional boundaries. Eg, quarter note triplets that start anywhere other than on beat 1 or 3, or an eighth note triplet that doesn't start on a beat. The steps schepers outline above is an example of this - if it worked, it would have created an eighth note triplet starting on the third part of another triplet. That probably should simply have been disallowed. We don't tend to do these things pn purpose, but they happen accidentally. My suggestion here: slow down, and don't make mistakes when entering triplets. If you do, use Undo liberally to get back to a known good state, rather than hacking away at it trying to correct the error. And again, if Undo fails to do exactly what it should, dlete the measure and insert a new one.

thanks for the advice, marc. in regards to #1, I actually saved repeatedly while working on the file. that's why i'm confused the backup is from 5 days ago when i first created the file. What i didn't do was close out of the file - I worked on it for many hours and saved often throughout. then i closed it. i reopened it about 20 minutes later and it was corrupt.. and the only backup version i had was from 5 days ago.

do you have to exit a file (or the program) for a new backup version to be created?

regards to #4, I did do a lot of copying & pasting of triplets into the middle of measures - because the score is chock full of triplets and they're all similar.

i suppose i'll start a new file and try to recreate - copying and pasting only complete measures that aren't corrupted. i'll follow your advice and save, close, and reopen often to see if it's working. i have a feeling i might be wasting my time but i don't see any other alternative.

r4902 of the stable branch contains a patch which allows to read the corrupted file.
I am further investigating the reason of the corruption...

hi - when i click on the attachment link i get a page that says "The requested page could not be found." Am I doing something wrong?

Thank you and werner both so much for your help!

lasconic, i don't know what you did but you're my hero, it's perfect. thank you so much.

Is there any advice you can offer for how to avoid a similar situation going forward? the score will a ton of triplets. don't copy & paste partial measures?

regardless i really appreciate you fixing the fie for me.