This can happen if there is an error very early during publication.

we don't support negative quantity (it is better represented by an
inversion of source and destination), and this allows better support of
the smallest slice in ERP5, more especially in SupplyLine_asCellRange,
where the title of the cell is computed with the min and max being None
or not.

See nexedi/slapos.core!307 (b7d5041e)

The real issue is probably in the launcher.js file, but it is more tricky to fix

See nexedi/slapos.core!313 (b02b2bff)

These are clone of Computer and Computer Partition, but used for virtual
computing devices, unrelated to physical hardware.

- Remove passing of **kw to query, since it would break Listbox
pagination
- Avoid code duplication
- Fix indentation

See merge request nexedi/erp5!1463

Otherwise, the object could be modified (_p_changed=1) and reindexed for nothing.

See merge request nexedi/erp5!1462

See merge request nexedi/erp5!1483

query SlapOs Master again and again if it created accordingly the Computer
Partitions as anyway scalability tests will start only when instance's state is
reported to SlapOs Master as ready. Thus in this case this check (and requests
to SlapOs Master) is not providing any valuable information at all.

See merge request nexedi/erp5!1482

Increase instance creation time from 10 to 15 minutes as instance creation time includes its frontend creation which can take sometime a longer than usual time.

See merge request nexedi/erp5!1481

In practice, SFTP-based client processes often use a prefix based
filter when retrieving files on the server.

ie: there is one SFTP server between 2 systems, used for different
exchanges. The SFTP client will do something like (found in the wild):
process1(file_list=sftp.list(path, prefix='PROCESS_1'))
process2(file_list=sftp.list(path, prefix='PROCESS_2'))

Our implementation of temporary_file_on_write only sets a suffix,
which doesn't work for the previous SFTP client. So let's add also a
prefix when writing temporary files.