Bug 876

There are some regressions / crashes / issues with running PDv3 decoders in PulseView that need to be investigated.

Here's an i2c->eeprom24xx example to reproduce this.

First, let's make sure the setup survives sigrok-test:

$ LD_LIBRARY_PATH=$HOME/sr/lib ./decoder/pdtest -r -v eeprom24xx/lcsoft-mini-board-fx2-init
eeprom24xx/lcsoft-mini-board-fx2-init/annotation ........................... OK
eeprom24xx/lcsoft-mini-board-fx2-init/binary/binary ........................ OK

This works fine, also many times in a row.

Now using sigrok-cli manually:

$ LD_LIBRARY_PATH=$HOME/sr/lib ~/sr/bin/sigrok-cli -i i2c/eeprom_24xx/atmel_at24c128/lcsoft-mini-board-fx2-init.sr -P i2c:scl=SCL:sda=SDA,eeprom24xx
Control code bits: 1010
Address bit 2: 0
Address bit 1: 0
Address bit 0: 0
R/W bit: read
Control word
Warning: STOP expected (not RESTART)
Data byte 00: FF
Data
Current address read: FF
Control code bits: 1010
Address bit 2: 0
Address bit 1: 0
Address bit 0: 0
R/W bit: write
Control word
Control code bits: 1010
Address bit 2: 0
Address bit 1: 0
Address bit 0: 0
R/W bit: read
Control word
Word address byte: 00
Word address
Data byte 00: FF
Data
Random access read (addr=00, 1 byte): FF

This works fine too, also multiple times in a row.

However, using PulseView to decode the same data from the same *.sr file causes crashes.

 - Sometimes it crashes upon adding the stacked eeprom24xx PD, other times that works.

 - Sometimes there are crashes when pressing "Run" multiple times. E.g. just add the i2c PD for that *.sr file, then press Run again and again. Eventually it will crash.


The issue could be in libsigrokdecode itself, or in PulseView, or a combination of both.

The fact that sigrok-test and sigrok-cli seem to work just fine might hint at a PV problem, but it could also be that their usage simply doesn't invoke the same code paths / problems as the PV use-case does.

Some potential things to look into: PV might feed in data / memory locations that are invalid (or not valid yet); there could be some double-free, stack overflow or similar issues in both repos; there could be thread related issues (both PV and also libsigrokdecode now use threads; in libsigrokdecode, each PD stack has its own worker thread).