Bug#1148127: at-spi2-core: FTBFS on i386, loong64, ppc64el: Direct connection pidfds accumulated after reconnecting

[Simon McVittie]

Source: at-spi2-core
Version: 2.62.0-1
Severity: serious
Tags: ftbfs
Justification: fails to build from source (but built successfully in the past)

The recent upload of at-spi2-core didn't build successfully on i386 with 
a test failure, making it non-co-installable on amd64+i386:

> 9/10 at-spi2-core:direct-connection FAIL             2.06s   killed by signal 6 SIGABRT
>――――――――――――――――――――――――――――――――――――― ✀  ―――――――――――――――――――――――――――――――――――――
>TAP version 14
># random seed: R02Sbecd00ba5c2fd03d825c0824cc26ff2b
>1..1
># Start of atk-bridge tests
>AT-SPI Received D-Bus name - org.a11y.atspi.test.DirectConnection
>not ok 1 /atk-bridge/direct-connection-is-released - Direct connection pidfds accumulated after reconnecting
># slow test /atk-bridge/direct-connection-is-released executed in 2.05 secs
>Bail out!

loong64 showed a similar test failure, so this doesn't seem to be a 
32-bit-specific problem. In the -ports ecosystem, so did alpha and 
sparc64.

ppc64el had a slightly different test failure that seems like it's 
probably related:

> 9/10 at-spi2-core:direct-connection FAIL             2.05s   killed by signal 6 SIGABRT
>――――――――――――――――――――――――――――――――――――― ✀  ―――――――――――――――――――――――――――――――――――――
>TAP version 14
># random seed: R02S376460e0a3bcbb7e4595fbdd8943bd7b
>1..1
># Start of atk-bridge tests
>AT-SPI Received D-Bus name - org.a11y.atspi.test.DirectConnection
>not ok 1 /atk-bridge/direct-connection-is-released - Direct connection pidfd was not released
># slow test /atk-bridge/direct-connection-is-released executed in 2.04 secs
>Bail out!

I've retried those three builds to see whether the failure is 
intermittent or reproducible.

    smcv