Re: error in info testsuite with dev/pts

[Gavin Smith]

On Mon, Feb 18, 2019 at 02:06:03PM +0100, Patrice Dumas wrote:
> There are many FAIL in info test suite, the most obvious error message
> in logs is (in general mangled mixed with output)
> cannot create dev/pts/15: Directory nonexistent

I made a change that was supposed to be a simplification, but it looks 
like it doesn't work.  The idea was that the test script and pseudotty 
could commmunicate via a single FIFO instead of two (one to get the name 
of the slave side of the pseudoterminal, the other to send bytes to be 
typed from the pseudoterminal).  However, it looks like pseudotty is 
reading back some of the bytes it wrote to this control channel.

> + read pts_device
> ./pseudotty: getting pty master fd
> ./pseudotty: unlocking slave device
> ./pseudotty: getting file name of slave device...
> ./pseudotty: /dev/pts/15
> ./pseudotty: opening slave device
> ./pseudotty: attempting to set window size
> ./pseudotty: ...succeeded
> ./pseudotty: entering main loop
> ./pseudotty: trying to read
> ./pseudotty: read byte 0x2F
> ./pseudotty: trying to read
> ./pseudotty: read byte 0x64
> ./pseudotty: trying to read
> ./pseudotty: read byte 0x76
> + exec
> + LIBC_FATAL_STDERR_=1
> + export LIBC_FATAL_STDERR
> + ginfo_pid=25429
> + killer_pid=25430
> + + printf q
> exec
> + timeout_test
> + wait 25429./t/malformed-split.sh: 175: ./t/malformed-split.sh: 
> cannot create e/pts/15: Directory nonexistent

Notice it was supposed to be /dev/pts/15 but it ends up as "e/pts/15".
The missing characters are accounted for by the Ox2F '/', 0x64 'd' and
0x76 ('v') bytes that pseudotty read.

I've reverted the change in commit 
53a14862a0ab7344cab5700ebb03ad9c9eeafd7f.