Will corrupted styles get fixed or we have to restore them by ourselves?
If you can send me some examples of corrupted styles, I'll see if there's a global fix. If it's going to require a fix per style, I'll probably have to ask you guys to take care of your own styles.
Glad to see you follow through. The couple I had been checking for reference seem complete now. I am still seeing a bug though. I updated my styles after the transfer, so I kind of expected them to revert to what they were in your backup. I haven't been updating Stylish with the code I'm uploading because I don't trust the process currently. Anyway, clicking "show css" on one of mine returns an error: https://userstyles.org/styles/133094/vk-aquatica-dark-redesigned-layout
Yeah, I can't speak to anything else, but that particular style seems back to normal. Can you confirm that you're not doing any more mass updates? If you still might, I'll wait to update styles.
HI Justin, I actually re-upped all of the missing ones last night so I'm good to go! also install counts are finally working too so you get a virtual high five from myself.
I've managed to partially resolve the search issues.
"Partially"? Which part works now, and which part doesn't?
An existing style is still isn't found after editing the metadata and updating, and it isn't found after then editing the code and saving. So what exactly does work?
Please wait with style updates until I confirm that things are stable again.
Confirm how, exactly? Three days ago you confirmed that the database was "secure, stable and backed up" [1] without saying how you did that, then admitted the next day that the styles database had been corrupted [2].
it looks like updating of styles doesn't work in Chrome and Opera. It says "Update completed", but code of the style remains old. Stylish is finding new updates over and over, but can't install them. In Firefox and Safari everything works well.
Maybe it's related to what hideheader mentioned earlier:
The "originalMd5" value in this style's Chrome file matches the Firefox file (df75db0d1aef277a24f8cf9c5af47002), though GOK why.
Stylish compares the local value of the checksum with the content of the .md5 file to determine whether an update is available. When a style is installed, Chrome uses the value in the .json file as the local value, and Firefox uses the value in the .md5 file. So, if the .md5 file isn't updated when the style is updated, Chrome will always think there is an update available even after you've installed it, and Firefox will never know there is an update.
All search related issues are now resolved. I'm aware about the Chrome and Opera upload issue, which as Hideheader said, is a known legacy issue. When I get the chance I'll review Jason's fix and push it to production. I want to try and resolve the counter problem first.
Searching by created date (descending) shows styles for 6 Oct but no later [screenshot 1], searching by weekly installs returns the order "0, 1, 0, 5, ..." [screenshot 2], and searching by keyword ("balrog") or domain ("trapperkeeper.com") for a style created after 6 Oct returns no results.
Added "whiskey tango foxtrot" to the description of an existing style ("My new style"), added /*whiskey tango foxtrot*/ to the code and wrapped the code in
@-moz-document domain(justinhindman.com) { ... }
and saved. Can't find it by "justinhindman.com", by "whiskey", "tango", "foxtrot", or any combination thereof.
Instead of saying "all search related issues are resolved", please say specifically what you have fixed. And yes, the installs count is a priority - it guides the consumers and feeds the producers. Making updates easier should wait until hell freezes over, for reasons that are obvious.
Searching by created date (descending) shows styles for 6 Oct but no later [screenshot 1], searching by weekly installs returns the order "0, 1, 0, 5, ..." [screenshot 2], and searching by keyword ("balrog") or domain ("trapperkeeper.com") for a style created after 6 Oct returns no results.
Added "whiskey tango foxtrot" to the description of an existing style ("My new style"), added /*whiskey tango foxtrot*/ to the code and wrapped the code in
@-moz-document domain(justinhindman.com) { ... }
and saved. Can't find it by "justinhindman.com", by "whiskey", "tango", "foxtrot", or any combination thereof.
Instead of saying "all search related issues are resolved", please say specifically what you have fixed. And yes, the installs count is a priority - it guides the consumers and feeds the producers. Making updates easier should wait until hell freezes over, for reasons that are obvious.
yeah we need the installs count like we need water and air.
the old version (with old description and title) which is already indexed, should show up in the search results, right?
No. Old version disappears from search results immediately. It takes a few hours for showing up new version in search. That's what I see with my styles.
Searching by created date (descending) shows styles for 6 Oct but no later [screenshot 1], searching by weekly installs returns the order "0, 1, 0, 5, ..." [screenshot 2], and searching by keyword ("balrog") or domain ("trapperkeeper.com") for a style created after 6 Oct returns no results.
Added "whiskey tango foxtrot" to the description of an existing style ("My new style"), added /*whiskey tango foxtrot*/ to the code and wrapped the code in
@-moz-document domain(justinhindman.com) { ... }
and saved. Can't find it by "justinhindman.com", by "whiskey", "tango", "foxtrot", or any combination thereof.
Instead of saying "all search related issues are resolved", please say specifically what you have fixed. And yes, the installs count is a priority - it guides the consumers and feeds the producers. Making updates easier should wait until hell freezes over, for reasons that are obvious.
yeah we need the installs count like we need water and air.
I found your test style on the second page when searching for 'whiskey' and on the first when searching for 'foxtrot'. Can you run another check and tell me if the problem still exists?
Regarding the search by descending created date, it's fixed now - the latest ones in the search results are from today.
Install counts are now getting updated, but I'm still going over it and verifying we're stable.
https://userstyles.org/styles/132545 Updated Oct 2, now (1) the .json file has no @document expressions and (2) the .md5 value doesn't match the .json "originalMd5" value.
Yeah, aware of this and am still working on it. Will update when I have more.
Guys, can you confirm you see an improvement in the search and weekly counter?
For me at least it seems to be working a lot better, I'm not sure if it's 100% accurate but it's at least going up an amount similar to what I'd been seeing before all this.
New problem - the Chrome .json files don't always match the Firefox .css files
[118959] [124745] [129961] The regexps are missing from the .json file, making them global styles in Chrome.
118959 updated May 30 is OK today; 124745 updated Oct 17 is OK today; 129961 updated Oct 05 is still missing 'Applies to' metadata, still has a bad .md5
118959 vs 129961 is troubling - it suggests that the backup used for the Oct 10/11 restore was corrupt. Styles updated between Sep 28/29 [1] and Oct 10/11 likely still have problems. (~400 styles)
... e.g. [128625] and [98423], both updated Oct 1. .md5 matches .json "originalMd5" which don't match the actual MD5 checksum for the .css file. Weird.
My theme 96784 doesn't work. 2/3 of code is missing.
The Firefox .css file is truncated to 16KB. The Chrome .json file is missing all but 11 bytes of the code. The style was last updated Aug 6; it should have been fixed on Oct 11 [1]. Apparently yours is not the only one [52077] [52079].
Your style should still be intact in the pre-Justin backup, but he's been very slow to fix problems and completely silent on what's been fixed, so you should re-upload it if you can instead of waiting.
Comments
An existing style is still isn't found after editing the metadata and updating, and it isn't found after then editing the code and saving. So what exactly does work? Confirm how, exactly? Three days ago you confirmed that the database was "secure, stable and backed up" [1] without saying how you did that, then admitted the next day that the styles database had been corrupted [2].
But the next problem is still there: Maybe it's related to what hideheader mentioned earlier:
Searching by created date (descending) shows styles for 6 Oct but no later [screenshot 1], searching by weekly installs returns the order "0, 1, 0, 5, ..." [screenshot 2], and searching by keyword ("balrog") or domain ("trapperkeeper.com") for a style created after 6 Oct returns no results.
Added "whiskey tango foxtrot" to the description of an existing style ("My new style"), added
/*whiskey tango foxtrot*/to the code and wrapped the code in@-moz-document domain(justinhindman.com) { ... }and saved. Can't find it by "justinhindman.com", by "whiskey", "tango", "foxtrot", or any combination thereof.Try to find this style, which I just created (16:42Z):
https://userstyles.org/styles/133756/13-october-2016
Instead of saying "all search related issues are resolved", please say specifically what you have fixed. And yes, the installs count is a priority - it guides the consumers and feeds the producers. Making updates easier should wait until hell freezes over, for reasons that are obvious.
Regarding the search by descending created date, it's fixed now - the latest ones in the search results are from today.
Install counts are now getting updated, but I'm still going over it and verifying we're stable.
for me only?
Good! Justin seem to be on a good trail to resolve the main issues. :bz
124745 updated Oct 17 is OK today;
129961 updated Oct 05 is still missing 'Applies to' metadata, still has a bad .md5
118959 vs 129961 is troubling - it suggests that the backup used for the Oct 10/11 restore was corrupt. Styles updated between Sep 28/29 [1] and Oct 10/11 likely still have problems. (~400 styles)
... e.g. [128625] and [98423], both updated Oct 1. .md5 matches .json "originalMd5" which don't match the actual MD5 checksum for the .css file. Weird. Still missing.
My theme 96784 doesn't work. 2/3 of code is missing.
I have to wait or this is definitely lost ?
Your style should still be intact in the pre-Justin backup, but he's been very slow to fix problems and completely silent on what's been fixed, so you should re-upload it if you can instead of waiting.