Subtle changes, bugs and feedback

That does help.

I hadn't yet turned back on the thing I had turned off.

Oooh... I think it may be large images and/or WEBP images.

What iPad mini do you have? (model, version, etc)

This could possibly be a Cloudflare bug and if so I'll get one of these devices acquired so that we can prove it one way or the other.

Ah...! I've just PM'd you the details. It does indeed seem to be on image heavy threads, sort of freezes then crashes while trying to load them. Don't go out of your way in acquiring things, I'm sure I can cope with the phone!

He works for Cloudflare...

And the suspected bug is webp.

We auto-convert JPEG images attached to things on the forum to WEBP if the user agent declares that it can support it.

It may be, that the iPad browser is declaring this... but can't actually support it.

If that is the case I could get a bit added to the server logic to skip webp for older ipads.

Just noticed that when looking at events, the +/- buttons to zoom the location map are shown over the drop down from the menu above. This also happens when browsing the Islington CC forum (where I first noticed).

I am using Google Chrome Version 58.0.3029.110, on Apple OS X 10.11.6.

Image attached in case I'm not clear.

1 Attachment

I can't fix that. It's weird IFRAME widget behaviour.

I've not tested this too extensively, but I noticed it while searching for posts by a certain author.

When I search for "authorId: xxxxx" I get all posts by that user. But when I search for "authorId: xxxxx "somestring"" I get a server error and my account gets logged out for some reason. Seems to happen for multiple strings and users that I've tried.

Same as this I guess:
https://www.lfgss.com/comments/13689061/

Ah yes! Seems like it.

Fair enough. Thought I'd mention it just in case...

Bossman - did you release something new this morning? I hit the forum downtime before 7, and now page load times feel improved, especially on sub 4G bandwidth.

Just a VACUUM FULL on the database server.

I do it periodically to free up space on the internally fragmented database files.

Yes a Noticeable loading speed up this morning

There's a bug that's been bothering me for a while. I used to be able to search a user's comments for a particular word, which was very helpful as you could find when someone had been posting about something but you couldn't remember where.

The way I'd do this is go to someone's profile, click comments, then type a word into the search box with 'comments' ticked under search options.

So if I wanted to find posts by @Velocio about Dunwich I'd end up here:
https://www.lfgss.com/search/?type=comment&authorId=47686

Then type 'Dunwich' into the search box.

This used to work, now it returns a server error. Is it fixable please?

Thanks for describing it.

Fixed.

Also... is this something you do frequently?

Worth having a search box on a profile page?

Always keep the carpets dust free :p

This would be v useful

@Velocio Out of professional curiosity, how large are the tables you're running? We're using postgres but I am not quite clear of what the upper limits are. I am sure you have loads more data than us. Do you shard it etc?

When I try to download the image of this post, I can only download it as .webp file. Although when I open it in a new tab, its url ends in .jpg. Same thing when I try to download it though. Why's that?

Using Chrome on Windows 10.

The tables are tiny, non-sharded... low tens of millions of rows per table. Absolutely nothing special being done with them.

I do know sharding techniques, and partitioning stuff... but I've not yet seen any reason to use those things here. The database is still less than 100GB in size.

It's because you're downloading using Chrome, and Chrome understands Webp, and Webp is about 1/3 smaller than the Jpeg.

Download using a non-webp aware browser, or something like curl or wget, and you'll not get the webp magic, you'll get a Jpeg.

Note that you may not get the original Jpeg... we compress those again too. But you'll get a Jpeg.

I'm getting multiple reload errors on the Digital Photography thread using iOS on my phone/iPad... doesn't really like loading at all at present.

I think iOS has some major issues with large images.

Since increasing the attachment size from 3MB to 30MB the threads in which people attach larger images have seen multiple people say that their iOS devices are either crashing, exiting the web browser or causing other issues.

I keep checking the pages and the only thing that they all have in common is that they have attachments of larger images.

I cannot test in iOS because I own no Apple things at all. I do need someone with iOS devices and who is a developer/debugger person to basically investigate further and see whether there is a specific asset that triggers an issue (was it an attachment or an inline image), or a specific URL (can be reproduced on non-image heavy threads, etc).

Until then, I'm afraid I have not much to offer other than to try an alternative device (not iOS) or avoid image heavy threads.

And do you run the cluster yourself etc, or managed solution like RDS? What kind of set up are we talking?
Also interested in what kind of approach you're using towards fulltext search?

We are running RDS postgres flavour ... I have no idea what's considered a large vs a small DB.
I really like postgres, especially the jsonb stuff, and have learnt a lot about it in the last two years. If there's a complicated reporting query they want, then it's always me writing them.

For fulltext right now I am running a trigger that processes the relevant fields into a ts_vector. Not sure if that's a good or shit approach.