We seem to have skipped a few versions, so an update would be good. I'll post an ebuild if I can (I'm currently running from git head.) Would there also be any point in creating a -9999 ebuild?
Created attachment 538290 [details] proposed ebuild This is simply copied from 2.5.3-r1. My preference would be to re-add use flags for gtkspell and gspell, requiring 0 or 1 of them, using the "internal" spell checker if neither is specified. (2.5.2-r1 allowed gtkspell, I don't know why it would have been dropped for using only gspell for 2.5.3-r1)
I'm temporarily marking this "NEEDINFO" since commit e6a952e8 (between 2.5.5 and 2.5.6) Mitigate EFail vulnerability for Webkit2 seems to have caused KMM to not display any HTML content at all. The person who filed that patch does still see HTML content, but is using a newer version of webkit-gtk then currently available in portage (2.18.6). Once this issue is resolved, we can decide whether to include a patch for 2.5.6 or just wait for 2.5.7 (or perhaps a 2.5.6.1). Note this is all with webkit-gtk. Using gtkhtml for HTML display in Balsa also seems to not currently work, but that is not new, will be addressed separately, and does not affect Gentoo, as it is not currently an available option in the ebuild.
I will add a 2.5.6 with patches to html.c for older webkit-gtk. Please test and confirm it works - I don't actually have any IMAP configured for it (nor do I really want to risk my real IMAP accounts to it), so stuck in the account configure wizard. gspell is used over of gtkspell because the latter is rather unmaintained now, with gspell being a sort-of successor to it. I'm not sure if internal spellchecker is worth it, as it still depends on enchant, just like gspell, plus probably more extra code to do what gspell does for it. Either way, I can't add it now, as enchant-2 is still masked and the necessary older <=1.6.0 is gone from tree by now. Feel free to explain why internal is better than gspell in a new bug and make it depend or block (as appropriate) the enchant-2 migration bug or something. I don't consider "it avoids an extra dependency" as a good reason, especially if "internal" means more compiled code inside balsa itself to sort-of achieve what gspell does. Review of the other changes welcome as well, or just a full review of the deps again (just doing changes according to build system changes diff in the bump)
I wonder if we should consider this a security bump and treat it as such, given your mention of EFAIL about the webkit changes...
The bug has been closed via the following commit(s): https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=650afd4647ec44b71457fa836c555f30ede718e9 commit 650afd4647ec44b71457fa836c555f30ede718e9 Author: Mart Raudsepp <leio@gentoo.org> AuthorDate: 2018-07-15 15:06:35 +0000 Commit: Mart Raudsepp <leio@gentoo.org> CommitDate: 2018-07-15 15:19:08 +0000 mail-client/balsa: bump to 2.5.6, includes EFAIL security fixes *html2text is actually a runtime dependency in the webkit-gtk code paths, not buildtime for all USE. * Add glib-compile-resources deps (libxml2). * New gnutls dep. * libesmtp usage removed. * Custom icon cache in /usr/share/balsa, needing gtk-update-icon-cache during compile phase. * src_prepare modifications are all obsolete (upstream bugs fixed). * Include patches to hopefully keep USE=webkit working with older webkit-gtk. Closes: https://bugs.gentoo.org/660322 Package-Manager: Portage-2.3.40, Repoman-2.3.9 mail-client/balsa/Manifest | 1 + mail-client/balsa/balsa-2.5.6.ebuild | 71 ++++++++++ .../files/balsa-2.5.6-fix-older-webkit1.patch | 156 +++++++++++++++++++++ .../files/balsa-2.5.6-fix-older-webkit2.patch | 53 +++++++ 4 files changed, 281 insertions(+)
The bug has been referenced in the following commit(s): https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=57635aef1d421e126ac2c733d9830e3c78899739 commit 57635aef1d421e126ac2c733d9830e3c78899739 Author: Mart Raudsepp <leio@gentoo.org> AuthorDate: 2018-07-15 15:34:33 +0000 Commit: Mart Raudsepp <leio@gentoo.org> CommitDate: 2018-07-15 15:40:34 +0000 mail-client/balsa: fix html2text handling for HTML mail quoting configure first checks for python-html2text binary, for which we had a dev-python/html2text dependency. But our pyhtml2text package installs a pyhtml2text binary instead, thus this will not satisfy the check, and HTML quoting support via a html2text binary gets disabled. Instead of tring to make it work with pyhtml2text, just depend on app-misc/html2text instead, which as a bonus has alpha keywords. This will make the balsa check pass with /usr/bin/html2text and HTML quoting should work (at least however good html2text over pyhtml2text is). Bug: https://bugs.gentoo.org/660322 Bug: https://bugs.gentoo.org/626690 Package-Manager: Portage-2.3.40, Repoman-2.3.9 mail-client/balsa/{balsa-2.5.6.ebuild => balsa-2.5.6-r1.ebuild} | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
Thanks for this. The ebuild works for me, using webkit-gtk-2.18.6, including showing HTML and some IMAP access (although I generally only use POP3). I do have some IMAP hiccups with gmail, but I get the same using git head, so I blame gmail, not the ebuild. I'm OK with the spell check issue. Note that I have gtk-spell 2.0.16 and 3.0.9 both using enchant 2.1.1. I'll be glad to post the (short) patches I used, if there is any interest. (I can't remember if I already posted them to the enchant bug, or if that's where I got them.) I'm guessing those patches could also be used for the internal spell-checker, but I accept your reasons why not to bother.
OK, thanks for testing. Regarding enchant, for me that's mostly just a case of "lots of other things are behind and need attention earlier", so no point in looking deeply for balsa yet. A dedicated bug for balsa enchant matters could be useful to remember about it once I do get to working on enchant-2 (unmasking, etc).