mixing of i686 and i386 archs

[andrlos]

It seems that a common practice in rpms is to mix i386 things into i686 rpms. This messes up with our tests, as the tests expect i686 folder but for example in java-1.8.0-openjdk-devel-1.8.0.292.b10-3.fc33.i686.rpm there is /usr/lib/jvm/java-1.8.0-openjdk-1.8.0.292.b10-3.fc33.i386/bin/ bin folder..
this results in rather large amount of tests failing as those are unable to locate relevant binaries in expected folders.. question is whether its something that should be fixed in the rpms or should be expected in the test.

[judovana]

the i686 and i386 should be synonyms from view of JSF.
Actually from view of jsf i686=i586=i486=i386==i286=x86-x86_32=86_32=86=intel32b
I recomend to have a better comapring function considering above equation, or to have some sanitize call before each operation unifying the 32b namings

[judovana]

maybe on el8 and up and fedoras on jdk11 and up, we can igore that some 32b java does exists at all...

[andrlos]

ok, will be done :-)

[andrlos]

this affects particularly these tests:
BinariesTest#_perform_all_checks
BinariesTest#all_jre_in_sdk_check
BinariesTest#check_java_cgi
BinariesTest#check_subdirectory_slaves
ManpageTests#man_page_test
ManpageTests#manpage_file_check
ManpageTests#manpage_links_check
PostinstallScriptTest#_check_post_in_script