[Info-vax] Backup API mailbox oddity

[Stephen Hoffman]

On 2018-01-10 22:58:31 +0000, osuvman50 at gmail.com said:

> On Wednesday, January 10, 2018 at 4:25:38 PM UTC-5, Stephen Hoffman wrote:
>> On 2018-01-10 20:39:23 +0000, osuvman50 at gmail.com said:
>> 
>> For this case, post some code, and somebody here can have a look.
> 
> I shared the code (353 lines) on my google drive:
> 
> https://drive.google.com/drive/folders/1egrNe1ZnAnRIl87Jw26sf9O9qs93OVYV?usp=sharing 
> 
> 
> The module that builds the rd_dirlist structure is omitted, two thirds 
> of the code in dir_sync.c is setup for the BACKUP$START() call and 
> mailbox I/O operations.
> 
> I could skip the mailbox I/O by building the backup control structure 
> dynamically (with potentially several hundred input files), but 
> concatenating control blocks of different sizes didn't appeal to me.

I took a very quick look at the code as a clear-my-head break from some 
other ongoing debugging work that's been causing serious headaches 
around here.

Based on what I see here, I'd just spawn BACKUP directly and capture 
errors via the termination mailbox.

I don't know that the BACKUP API was ever tested with this 
mailbox-as-input approach, and it certainly appears that it's not 
working or not working very well.

If you want to maintain the use of that that mailbox as a way to pass 
in the list of files, then I'd try specifying that mailbox on the 
BACKUP DCL command line as the first parameter.  Use a leading @.   I 
suspect a spawn will involve less source code, too.

FWIW, I'd only use $hiber with a $schdwk previously requested, as it's 
far too easy to get stuck in a $hiber otherwise.  Mentioned as I've 
seen apps stall in $hiber due to a missed wake or an unexpected wake 
coalescence.

That string upcase logic reminds me; I do need to rummage around for 
the current state of ICU support for OpenVMS and for u_strToUpper and 
other giblets.  Since there are system and RTL calls all over the place 
in that code, str$upcase is also an option.   Probably a safer approach 
is case-blind with strncasecmp or str$case_blind_compare.  Though none 
of that knows squat about UTF-8.  (I only mention any of this as 
upcasing and downcasing and string comparisons have nailed code in more 
than a few projects, and OpenVMS does have both UTF-8 and VTF-7 strings 
lurking in the file naming here, albeit only very obscurely documented.





-- 
Pure Personal Opinion | HoffmanLabs LLC