[Rpm-maint] Sigpipe handling in build/rpmfc.c

[Panu Matilainen]

On 06/06/2011 02:40 PM, Michael Schroeder wrote:
>
> Hi guys,
>
> what's the reason for the complex pipe handling with sigpipe_handler(),
> sigpipe_init() and sigpipe_finish()?

The point was to make it more robust:
a) use blocking io instead of spinning on a timeout that will inevitably 
be wrong for some case somewhere
b) be tolerant against errors, such as missing script interpreters and 
bad arguments which cause the generator helpers to exit before reading 
their stdin at all

Both being more important than before now that we're encouraging 
packagers to add dependency generators on their own instead of piling 
everything into rpm itself.

> IMHO aborting the read when a SIGCHLD occurs is wrong, as the
> pipe may still contain unread data (and data is only read in
> sizeof(buf)-1 chunks). Wouldn't it make more sense to simply
> ignore SIGCHLD and read until an EOF is returned, i.e. the
> read returns zero bytes?

SIGCHLD needs to be caught to avoid hanging in the select if the child 
exists without flushing its stdin, but you're probably right the reading 
side could (in some rather rare conditions) miss some final bits from 
the child output. Hum, I guess SIGCHLD should only abort the loop if 
there are unwritten bytes left. Need to scratch head + consume coffee 
some more...

Mind you, I do agree the thing is more complex and error prone than I 
care for. On more than one occasion I've considered replacing it with a 
dirt-simple setup using temporary files and letting the shell deal with 
the pesky little details.

	- Panu -