Compiler and platform tuple

[MBlesken]

The current specification has lacks and inaccurate references to compiler and platforms:

A few examples

• The possible values of the attribute compiler of SourceFileSet is not precisely defined. Only examples are given. (An importer has to identify the compilers it supports.)
• The Application Binary Interfaces (ABI) listed for the platformTuple intended to define the directory for binaries is likely incomplete and lacks of concrete descriptions (E.g. "gnu": If sources are compiled for linux elf format is created. If you compile for windows the pe format is created. But pe is not listed. So should gnu be used here?)
• The ABI Versions only lists MSVC

The question is: Is it intended to describe these details in the specification or can we provide a concrete reference an FMU developer has to regard?

[chrbertsch]

Must these definitions be in the FMI3.0 standard, or could we reference e.g. to a website, where we could update it independent of the release cycle of the FMI standard?

[andreas-junghanns]

@t-sommer : Please add examples for the most common cases for source, static binary and dynamic binary.

[MBlesken]

The current specification has lacks and inaccurate references to compiler and platforms:

A few examples

• The possible values of the attribute compiler of SourceFileSet [NOW: -> BuildConfiguration, PR Build configuration compiler #1465] is not precisely defined. Only examples are given. (An importer has to identify the compilers it supports.)
• The Application Binary Interfaces (ABI) listed for the platformTuple intended to define the directory for binaries is likely incomplete and lacks of concrete descriptions (E.g. "gnu": If sources are compiled for linux elf format is created. If you compile for windows the pe format is created. But pe is not listed. So should gnu be used here?)
• The ABI Versions only lists MSVC

The question is: Is it intended to describe these details in the specification or can we provide a concrete reference an FMU developer has to regard?

@t-sommer @andreas-junghanns @chrbertsch These parts of the specification have not been addressed by PR #1316. (My PR #1465 isn't addressing it either.) Could you please comment on the issues I marked? I'm not an expert and also would need support here, but from what I got when discussing with colleges is that these parts of the standard are not helpful with the intended workflow.

[andreas-junghanns]

@t-sommer @andreas-junghanns @chrbertsch These parts of the specification have not been addressed by PR #1316. (My PR #1465 isn't addressing it either.) Could you please comment on the issues I marked? I'm not an expert and also would need support here, but from what I got when discussing with colleges is that these parts of the standard are not helpful with the intended workflow.

@MBlesken : Could you work on a proposal with your colleagues?

[andreas-junghanns]

Both related PRs where closed. @MBlesken can we close this issue as well?

[MBlesken]

To my understanding the current state of the standard contains an unsuccessful attempt to enable an FMU generator to provide all the information that an importer would need to decide whether it can or cannot support a static library contained in the FMU. From the discussions with my colleagues I concluded that in general this information can in general not be provided or would be extremely complex and thus not practical. (For now we thus decided not to support static libs.) As a consequence the FMI desing group should first of all decide what the goal should be i.e. which workflow should be enabled. Afterwards we should either cut back the current attempt, clean it up or and extend it. - @andreas-junghanns @chrbertsch @t-sommer what is your opinion on how we should proceed?

[andreas-junghanns]

In a sense I am reminded to the 2.0 source-code file description: It helped a bit, but was not complete.
And I agree with @chrbertsch : If we could simply reference another standard, we could solve this neatly.
Any ideas where to look?

Less ideal is a list of common and mostly used examples and possible a heuristic how to build the description names for those not listed. @t-sommer added those in 2.5.1.4.2. Platform Tuple Examples. @MBlesken, did you see that recent addition? Is this not enough for a) most commonly used or b) as heuristic description on how to build these names?

I propose to find a slot today to discuss this.

[andreas-junghanns]

@MBlesken : we close this because #1501 was merged.