[Monotone-debian] Bug#624779: Bug#624779: monotone: FTBFS on hurd-i386

[Stephen Leake]

Pino Toscano <address@hidden> writes:

> Package: monotone
> Version: 1.0-2
> Severity: important
> Tags: patch
> User: address@hidden
> Usertags: hurd
>
> Hi,
>
> currently[1] monotone fails to build on hurd-i386.
> The problem is due to the usage of MAXPATHLEN, which is not defined on Hurd
> (as there is no length contraint for paths).

Ok.

> The attached patch does the following fixes:
> - src/netxx/serverbase.cxx, Netxx::ServerBase::bind_to():
>   use glibc's get_current_dir_name() if no PATH_MAX (the correct define to
>   eventually use) is defined, or use the current getcwd() way; in both the
>   cases, the cwd read is done just once at the beginning of the function,
>   instead of being done everytime in the for loop

This looks reasonable.

>   (also, couldn't this file eventually use get_current_working_dir()
>   from fs?)

That would be cleaner. Originally, netxx was a third-party library that
monotone was using, so we were reluctant to make many changes in it.
That is no longer true, so we might as well use get_current_working_dir
now.

> - src/unix/fs.cc, get_current_working_dir():
>   use the same working patters as above, making use of PATH_MAX instead of
>   hardcoding 4096 as size

Ok.

> Also, after the (log!) test suite run I got just three failures:
>  60 automate_certs                                FAIL (line 2) 0:09,
>  0:00 on CPU

This is used by front-ends to create branches, so it's not an acceptable
failure. Can you post monotone/test/work/func.log (that shows the
complete test run log for the failure cases).

> 177 database_dump_load                            FAIL (line 25) 1:08, 0:00 
> on CPU
>   2 bash_completion                               FAIL (line 43) 0:49,
>   0:00 on CPU

These failures are more acceptable, but they also might be easy to fix.

-- 
-- Stephe