Other bugs found in DGS1 Gnote files
Jump to navigation
Jump to search
This information comes from looking in Gnote files on DGS1
More than likely these are still mostly open issues.
Follow up: spreadsheet woes and CPLD registers
A couple months ago we tried to update spreadsheets and much evil ensued. The whole EPICS database went crap and a couple of highly worrisome days of reverting and rebooting occurred. In the hopes of forestalling another round of that, here are some observations:
1) There is evidence that using an EPICS PV type of 'T' (toggle) may be deleterious to your health. Avoid.
2) There is evidence that adding any register to a spreadsheet that has an address greater than 0x7FFF is also deleterious to your health. Avoid.
3) There is circumstantial evidence that having white space in column 'P' of a spreadsheet (selections for type 'b' or 'm' PVs) may also be problematic. It is suggested that this also be avoided.
4) If you are defining a 'b' or 'm' field, do NOT use colons in the text name of the setting. This mucks the Python parser and generates malformed PV database entries. Example:
0:text:0;1:other text:1 is ok.
0:text:0;1:other : text:1 will fail.
No error given by Python, but if you look carefully in the IOC console there's a error there and you can find malformed definitions in the .template files.
There are also some IOC idiosyncrasies that should be noted:
A) After going off to con5 and doing a new make for the system, simply rebooting the TrigCPU (VME32) isn't sufficient to get the system to re-initialize. You have to go to the crate and cycle the power, or hit the crate's reset button, to get the other IOCs to understand things have changed. It appears that only the hard loss of the Ethernet connection caused by a hard boot of the trigCPU IOC suffices to get all the other IOCs to reload the database.
B) For reasons nobody understands, some of the digitizer IOCs are a lot slower than others. If you power cycle the crate containing VME32 (trigger), VME9 and VME10, for example, the trigger comes up in the time expected but VME9 and VME10 aren't ready for a few minutes after that.