From mboxrd@z Thu Jan 1 00:00:00 1970 From: Peter Korsgaard Date: Sat, 02 Sep 2017 01:18:52 +0200 Subject: [Buildroot] [PATCH 1/1] package/transmission: fix gtk support In-Reply-To: <5fe9c108-7a04-ca7c-9bd8-aed7ac639026@mind.be> (Arnout Vandecappelle's message of "Wed, 12 Jul 2017 12:04:42 +0200") References: <20170710175137.1402-1-bernd.kuhls@t-online.de> <87a84asr2s.fsf@dell.be.48ers.dk> <20170711231701.3850d479@windsurf.lan> <20170712094924.435a5c54@windsurf.lan> <5fe9c108-7a04-ca7c-9bd8-aed7ac639026@mind.be> Message-ID: <873786qa77.fsf@dell.be.48ers.dk> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: buildroot@busybox.net >>>>> "Arnout" == Arnout Vandecappelle writes: Hi, >> Why should we randomly pass --enable-nls to some packages? Why should >> this package depends on BR2_SYSTEM_ENABLE_NLS, if it needs NLS support? > Because BR2_SYSTEM_ENABLE_NLS does not really say that NLS *support* is > enabled, it says that NLS *installation* is enabled. NLS support is always > "enabled" through the stubs in musl/uClibc. > As far as I understand, the configure error is just there because the > transmission guys are too lazy to have conditional use of the intl functions. I > don't see how there could be a hard requirement on non-stub implementations of > the intl functions - except if they start calling internals. So the better > solution, actually, would be to patch out this check in configure and instead > check for intl and error out if that is not available, independent on the > --enable-nls option - just like most packages do, I think. > If you're not actually interested in translations, there is no reason really > why you should pull them in just because this package is calling gettext stuff > unconditionally. > That said, probably nobody really cares, we're just fixing build failures here. > So in the end I don't really mind if we depend on BR2_SYSTEM_ENABLE_NLS as the > easy way out. Though then the commit message should at least provide the full > story, as a hint for future contributors who want to fix it properly. Correct. I went with depending on BR2_SYSTEM_ENABLE_NLS, but I added a comment explaining that we could also explictly pass --enable-nls. -- Bye, Peter Korsgaard