search

larshp / abapGitServer

Git server implemented in ABAP
MIT License
64 stars 25 forks source link

installation fails with multiple inactive objects #170

Closed AndreaBorgia-Abo closed 1 year ago

AndreaBorgia-Abo commented 1 year ago

I am trying to install AGS on 3 different systems, called SR2, SR3 and S3E, SAP_ABA versions are as follows: SR2: SAPKA73105 SR3: SAPKA74004 S3E: SAPK-75006INSAPABA

I have:

These are the screenshots of the error messages, system name is visible in each one: grafik grafik grafik

In each case, data element ZAGS_REPO was partly active but could be activated manually. More interesting, table ZAGS_MERGE_REQ could not be activated because of the CREATED_BY field, having a nonexisting SYST_UNAME data element. My MiniSAP is currently boxed awaiting relocation, but a colleague has verified that it exists in newer HANA versions so I am going to fork and check if I can fix it with a custom element.

larshp commented 1 year ago

thanks, pull requests welcome

larshp commented 1 year ago

use https://abapedia.org/open-abap-core/syuname.dtel.html if possible its marked as released in Steampunk

AndreaBorgia-Abo commented 1 year ago

SYST_UNAME is not the end of the story: I was wrong, S3E has it but only after changing to SYUNAME could I pull the repo onto this system. S3E still has a couple of spurious diffs related to service definitions but the service apparently works: grafik

Conversely, SR2 still barfs when activating objects, even from my "syuname" branch. grafik

Tomorrow, we'll see how to deal with this.

AndreaBorgia-Abo commented 1 year ago

On SR2, the oldest of the bunch, fm ZAGS_HANDLE_MERGE_REQUESTS must be rewritten to avoid inline declarations. (for the sake of consistency, I renamed the syuname branch to backcompat. Since there are possibly many such cases, maybe I should trigger the linter via ATC and find them all, we'll see.

larshp commented 1 year ago

yea, adjust https://github.com/larshp/abapGitServer/blob/main/abaplint.json#L27 via a pull request

AndreaBorgia-Abo commented 1 year ago

Done: #172

mbtools commented 1 year ago

@AndreaBorgia-Abo, your fork is not up-to-date.

in the SR2 log, I can see that table zags_tree_cache has index z10. however, Lars' repo does not contain this index.

AndreaBorgia-Abo commented 1 year ago

@mbtools could it be because I merged the silly own PR I created by mistake? I'll have a look at the table, I have no idea where that index comes from EDIT: the table looks fine in the repo, in the system I have manually activated it and the index panel is empty. Where does aG handle table / index activation?

mbtools commented 1 year ago

Right. Your repo is fine, so the index was just local.

dd12v in the tabl.xml: https://github.com/abapGit/abapGit/blob/main/src/objects/zcl_abapgit_object_tabl.clas.abap#L365

AndreaBorgia-Abo commented 1 year ago

This is getting weirder / more intriguing by the minute: grafik May I clear these entries the hard way? They weren't there before yesterday, so it's not like I'd be removing something required for other users. That way, I could really start from scratch and see where they get written.

AndreaBorgia-Abo commented 1 year ago

After removing Z10 from the internal table in debug (not yet from the DB table), I get to this point: grafik

mbtools commented 1 year ago

ECTD: https://github.com/larshp/abapGitServer/issues/154 don't think this was resolved properly

SMIM: https://github.com/abapGit/abapGit/issues/4783

AndreaBorgia-Abo commented 1 year ago

Ok, then I'll pause work on SR2 (oldest system), quickly recheck if SR3 also has index issues and then be on my way with tests on the newer functioning AGS installation. DD12V on SR3: grafik After forcing the index removal in debug as before, I get to the same eCATT Help error on SR3 as on SR2, so I'm done with them for the moment.

mbtools commented 1 year ago

there are a couple more places that reference syst_uname. one is in the ectd object:

https://github.com/larshp/abapGitServer/search?q=syst_uname

suggest changing those two as part of #171. it might fix the ecatt error, too

AndreaBorgia-Abo commented 1 year ago

Thanks, I hate it that forked repos are not searchable :/ However, I have already changed the first occurence and the second one by the looks of it is documentation.... still, I'll change it for the sake of consistency.

AndreaBorgia-Abo commented 1 year ago

Updating SR2 and SR3 again with the "backcompat" branch, the ECTD object can be updated and the only remaining gotcha is apparently this: grafik

SR3 now works (web interface comes up), while SR2 is still complaining about that SMIM object, I've tried pulling one object at a time and it works until you get to that one, then they all show up again as modified. For our purposes, that is more than enough.