On 03/25/10 02:57 PM, Shawn Walker wrote:
> On 03/25/10 01:52 PM, Joseph J VLcek wrote:
>> On 03/15/10 06:41 PM, Shawn Walker wrote:
>>> On 03/15/10 12:55 PM, Dave Miner wrote:
>>> ...
>>>> 370: I was surprised that this isn't encapsulated in an eval in
>>>> order to
>>>> get the transaction id into the environment; perhaps that's not
>>>> necessary for file: FMRI's? However, I'd expect it's necessary to
>>>> support http: FMRI's, which would be nice to allow, so does it break
>>>> anything to have the transaction id? Obviously, if we do support http:
>>>> FMRI's, then the comments in the manifests need updating.
>>>
>>> It's necessary regardless of what type of destination repository is
>>> used.
>>>
>>> Although, you could parse the output of "open" to get the transcation id
>>> and set the environment var yourself instead.
>>>
>>> pkgsend currently doesn't have a global command line option to provide
>>> the transaction id for every subcommand (oddly, only close does).
>>>
>>> Cheers,
>>
>> Shawn,
>>
>> The updated webrev can be found here:
>>
>>
>> http://cr.opensolaris.org/~joev/bug15163_r2/
>>
>> I have reworked this section of the code.
>
> This seems better to me, although again, I'm not a shell expert, so I'll
> defer to others on the precise shell syntax.
>
> The only other question I have is whether the user will see the exact
> failure output of pkgsend import and close or if they too need to have
> their output captured and passed to the error handler?
>
> Cheers,
Thank you Shawn.
Thank you for thinking of this. I have just confirmed the error output
from these commands will be written to both the log and the terminal
window where the distro constructor is running.
I confirmed this by forcing an error. For example I cobbing "GEORGE"
into the script as follows:
"pkgsend -s GEORGE import ${PKG_IMG_PATH}"
The output clearly indicates the problem:
...
==== ai-publish-pkg: Publish Package
pkgsend: Malformed URL: 'GEORGE'.
Invoking: error_handler
pkgsend import failed
...
Thanks again
Joe
