configobj 5.0.5 fails its test suite on uclibc. Reproducible: Always Steps to Reproduce: 1. FEATURES="test" emerge -1 dev-python/configobj 2. 3. Actual Results: emake failed (test phase) Expected Results: configobj updated on system Several other Python packages fail their tests on uclibc.
Created attachment 409064 [details] build.log
Created attachment 409066 [details] emerge --info output
Probably no one of us can test it. So uclibc guys, could you please assist?
All the errors have the following text in their message : "OSError: illegal IP address string passed to inet_aton" Looks like uclibc does not support x.y.z , x.y forms, nor a decimal string of a 32-bit integer in its inet_aton, probably for size considerations. This function is not defined in POSIX, but by 4.3BSD, however inet_addr, for which inet_aton is a replacement, is in POSIX-2001 and defined all these unusual forms and event support octal and hexadecimal (holy cow!). And yet inet_pton, which also supports IPv6 addresses and in POSIX, only supports the dotted-quad for IPv4 addresses. How should we handle this?
(In reply to René Rhéaume from comment #4) > All the errors have the following text in their message : > > "OSError: illegal IP address string passed to inet_aton" > > Looks like uclibc does not support x.y.z , x.y forms, nor a decimal string > of a 32-bit integer in its inet_aton, probably for size considerations. > This function is not defined in POSIX, but by 4.3BSD, however inet_addr, for > which inet_aton is a replacement, is in POSIX-2001 and defined all these > unusual forms and event support octal and hexadecimal (holy cow!). And yet > inet_pton, which also supports IPv6 addresses and in POSIX, only supports > the dotted-quad for IPv4 addresses. > > How should we handle this? Sounds to me like we port over what glibc or musl does and properly guard it with _USE_BSD. Its on the todo list.
sys-libs/uclibc has been removed from the tree, replaced by sys-libs/uclibc-ng. if this is still a problem on uclibc-ng, please open a new bug.
