From mboxrd@z Thu Jan 1 00:00:00 1970 Content-Type: multipart/mixed; boundary="===============4360931946275745930==" MIME-Version: 1.0 From: Philip Tricca Subject: Re: [tpm2] building tpm2-tools with bitbake fails even though tpm2-tss builds fine Date: Tue, 20 Feb 2018 13:49:07 -0800 Message-ID: <20180220214907.GA6093@intel.com> In-Reply-To: 2ba64c1cd60f46bda1192e4741b926ef@AZDC-MMB02.GD-MS.US List-ID: To: tpm2@lists.01.org --===============4360931946275745930== Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: quoted-printable On Tue, Feb 20, 2018 at 08:37:36PM +0000, Scheie, Peter M wrote: > I'm trying to build tpm2-tools with bitbake, but it fails saying it canno= t find sapi ("Package requirements (sapi >- 1.3.0 sapi < 2.0.0) were not me= t") which is provided by tpm2-tss (right?). Tpm2-tss builds just fine, and= when I try to build tpm2-tools, I can see tpm2-tss being built since it's = listed as a dependency in the recipe. But something breaks somewhere. This is bitrot: I've been neglecting these recipes for a while now. They're a personal project, not on my Intel time. Still, Trevor has been proding me along with some very helpful patches. We've got a 'rocko' branch where we're refreshing versions and accounting for minor changes in the upstream build. I think the TSS2 libraries and the daemon should now be building the latest release but I haven't done the tools yet. This could be part of what you're running into. > I should add that I'm not using the git clone of tpm2-tools, but rather t= he tar file from releases. For my recipe, I've modified tpm2-tools_2.1.0.b= b so that it points to the tar file rather than the github repo, and rename= d it tpm2-tools_3.0.3.bb. Working through this, I'm confused by several th= ings: > = > * Why is the recipe version 2.1.0 on the OpenEmbedded.org website= when the latest tar file is 3.0.3? See above. Also: https://github.com/flihp/meta-measured/pulls/ > = > * Why is the recipe made up of two files, tpm2-tools_2.1.0.bb an= d tpm2-tools.inc instead of just putting it all in the .bb file? I don't remember. I think I was trying to support multiple versions of one component at some point and the build required different work-arounds in the bitbake metadata. As the build has gotten better the version specific files (*.bb) have shrunk as the workarounds have come out. Eventually they may go away completely ... > * Why does the SRC_URI in the .bb file point to gcc7.diff which r= eferences, among others, a main.c file that isn't in the tar file? No clue. The right thing to do is to update the recipe to the latest and go from there. Philip > As I said, I've modified the recipe files to use the 3.0.3 tar file for t= he SRC_URI, but I suspect something else is not quite right perhaps because= of how I'm mixing versions of tar files and recipes (even though I've remo= ved all references to the 2.1.0 version) and causing the sapi error. Ideas? --===============4360931946275745930==--