Possible wrong buffer overrun detection during lzw_decode

[alimbourg]

... Causing last pixel of every lines of non interlaced gifs to be glitchy.

When activating LZWDEBUG (=1) , serial monitor is reporting a lot of buffer overruns on very small gifs (8x8) (LzwDecoder_Impl.h/line 121)

It seems that decompressAndDisplayFrame() is subtracting one byte for the buffend parameters ( GifDecoder_Impl.h/line 869) while calling lzw_decode (although the inner code code seems not require that unless I'm missing something).

Changing the call into int len = lzw_decode(imageBuf + tbiImageX, tbiWidth, imageBuf + maxGifWidth, align); suppress the warning and decode the last pixel. Regards

[prenticedavid]

Sorry, I have not really done anything with this project recently.

The master Branch is out of date. I had been using adafruit_opt Branch.

I should tidy up the comments in the code and put it into the master Branch.

Please can you quote which Branch and which GIF is exhibiting the problem.

David.

[alimbourg]

Sorry about that: so, previously tested on master, but the call to lzw_decode and the dubious bufend computation is similar in the adafruit_opt branch (GifDecoder_Impl.h/line 873)

Here is the tiny GIF showing a bizarre last pixel on each row when lzw_decode is leaving too early:

Regards

[prenticedavid]

Your Tiny gif is difficult to see on a PC screen. I copied it to the data directory of the AnimatedGIFS_SD sketch. Uploaded to SPIFFS

It was pretty small to see on a 135x240 ST7789 screen So I tried a 128x160 ST7735 screen where iit was still small.

Each frame on ESP32 seems to match the frame on my PC running EZGif

Ah-ha. I loaded the tiny.GIF into IrfanViiew. Frame #2 is 8x8 Blue. WxH Frame #3, 4, 5 seem to lose the last row i.e. 8x7 Red/Orange

I will investigate. Do you have a better example GIF ?

David.

[pmerlin]

thank you @alimbourg, I had the same issue : last pixel of each line was background. This is solving my issue (on 16x16 gif)