Re: [gentoo-dev] [PATCH 1/2] bzr.eclass: Reinstate eclass

[Ulrich Mueller]

>>>>> On Sat, 25 Sep 2021, Arthur Zamarin wrote:

> On 25/09/2021 12:36, Ulrich Müller wrote:
>> +		if [[ -z ${EBZR_INITIAL_URI} ]]; then
>> +			bzr_initial_fetch "${EBZR_REPO_URI}" "${branch_dir}"
>> +		else
>> +			# Workaround for faster initial download. This clones the
>> +			# branch from a fast server (which may be out of date), and
>> +			# subsequently pulls from the slow original repository.
>> +			bzr_initial_fetch "${EBZR_INITIAL_URI}" "${branch_dir}"
>> +			if [[ ${EBZR_REPO_URI} != "${EBZR_INITIAL_URI}" ]]; then
>> +				EBZR_UPDATE_CMD="${EBZR_UPDATE_CMD} --remember --overwrite" \
>> +					EBZR_OFFLINE="" \
>> +					bzr_update "${EBZR_REPO_URI}" "${branch_dir}"
>> +			fi

> This logic maybe can be improved and made easier, by defaulting
> EBZR_INITIAL_URI to be by default EBZR_REPO_URI (as a result, the
> first call to bzr_initial_fetch can be done outside the if), and if
> the ebuild writer changed it (check using the if here), then do the
> next bzr_update.

> Or another option, instead of setting by default, do the first fetch as:
> bzr_initial_fetch "${EBZR_INITIAL_URI:-${EBZR_REPO_URI}}" "${branch_dir}"
> But I think this one is less readable.

>> +		fi

I am going to drop support for EBZR_INITIAL_URI altogether. This was a
special case used for GNU Emacs, which historically had a slow original
repository and a fast (but possibly out of date) mirror. It is unlikely
that we will need this again.

Ulrich